📚┃suno-school

[frigid cove]

Yeah, I'm SUPER tempted to try, but I'd burn through credits so fast. 😂

I'm going to post a full song in a moment where I just combined the lyrics for 2 generation attempts using the prompt box to ask "how it feels" about making music for us. 😂 Only real editing I did was change the chorus in the 2nd bit to match the same chorus in the first.

I dunno, I just found it kinda interesting what it came up with.
https://suno.com/song/24112bba-0b67-45b5-b41f-45b521329e7d

[charred merlin]

"generate lyrics" doesn't use credits, afaik, only clicking "create" uses credits.

[frigid cove]

Oh. I meant like generating songs. XD I never considered trying to just generate lyrics... or knew that was a function. >_>

[charred merlin]

Apparently Suno's first name is "Melody", btw

[frigid cove]

why does that make too much sense

😂

[frigid cove]

That's... that's actually super awesome.

[alpine venture]

I'm not sure if anything like this exists on suno

https://suno.com/song/ac18e4d3-7ed6-4952-88d6-507c5a08d3b8

[alpine venture]

sl (minor fry): this is the "punchy"/edgy part of slacker, but it can be a little trite when used alone

ot (medium vocal fry, female bias ie mild data corruption) - melodic insightful build, bright generally, not quite happy, but builds the melody in a sort of bright insight direction

ica (pretty clean vocals, pop-like) - sad piano, very close to the inverse of ot

ello - almost mellow tones, usually creates like three note patterns or repeating arpegios, but overall this one is a light touch

umble (clean vocals, extraorinarily so): folksy weirdly, but also can go "hard"

My (very mild compression sound) - high variety, wide, generally mediocrity from about ten years ago

mys (good clarity, mild compression sound): mostly rock, trad rock id call it. Just a drum beat.

unk (good clarity) - bouncy, poppy

cr (high clarity, i believe vocal focused, has capacity for real quality melodic progression) - very style agnostic, although it does contain deep bass capacity, and has some wierd like, strange af voice/cloying

# (mild fry) : high pace, excited, strong pace, energetic, almost hyper-lite

hyper (mid fry): strong favor to female vocals, makes the song kind of just a normal modern pop song, but like, only the boring parts

anc - grungy

sw - bright, clear, lose sound

ore - (bad voice)

ina (slight gurgly) - good deep steady beat, very close to that hispanic beat that shows up on a lot of stuff

b (wild, has some 90s pop in there, but a really really ubique clean voice)

ural (gurgle) - 90s rock?

Mb (pretty clean) - highly generalized african music token; appears to influence variance ie ot creates dynamism? either that or "inal"

aka - a knifes edge token, as it is related to Baka

inal - secokd half to seminal, it seems to be highly "complete", insightful, baroque

sem - chaotic, dirty, good

my current incomplete note list

I have the executive function of a drugged chimpanzee, so basicslly I find a lot of stuff that I just don't remember to write down

NOTE
if you use these, you can generally combine them into new words and make new genres

however

you need to check with the tokenizer to confirm that the tolenization remained correct

I'm that config

sometimes it'll do weird things

one trick you can do is adding - before your tags

which will force the tokens to "set" to this way

[tight spade]

whats a tokenizer?

[alpine venture]

https://platform.openai.com/tokenizer

link above

it's how the model PROBABLY sees our inputs

not necessarily this one

but I'm p confident

based on how Mb is so well generalized

and breaks what had previously been a rule I'd observed where suffixes are less "sharp" than prefixes

which has to do with how much more common they typically are

however, Mb is in SO much african music

tagging data

and so it got picked up in a way that allows you to take core concept from it

like pilyphony

and generalize them

[charred merlin]

Ai doesn't understand words the same as we do, it needs to break words down into "tokens" -- little bite-sized chunks of ideas.

You can use the tokenizer to see how an AI would split your input into "tokens" that it can understand.

Most AIs can only understand a certain number of tokens at once before they start messing things up.

[tight spade]

ahhh makes sense

and what about suno how many till it explodes

[alpine venture]

Anyway, the prompt on the most recent song

sem is part of seminal

but alsoclikely semitic

[tight spade]

ok i get now

[alpine venture]

although I didn't test that

yea, sorry, just making notes for anyone who wants to use the info

[tight spade]

i intrested'

[alpine venture]

I think it's good for people to share, helps improve all the music

and I like punishing my ears sometimes, but I literally have to throw my headphones sometimes if I let a certain playlist go by itself

Anywho

sem - part of seminal which is powerfully involved in classical composition

however inal is waaay more "baroque" and structured by comparison

[charred merlin]

The HARD limit is how many characters you can use in the lyrics / style box.

The closer you are to FILLING it, the worse it CAN get. Not guaranteed to be bad if you fill it, but chances get higher.

[alpine venture]

in any case, I iterative experimented, but like a sampler if anyone has watched an image generation, and sort of hunted for rhe sound

[tight spade]

ahh ic "says drax" lol

so useing the web.site you listed how do i impute that data into suno manually?

or could i have it run in gpt to out forme?

[charred merlin]

You misunderstand, you do not use that website to input to suno

[tight spade]

i under stand that you use it too shorten words? this alot to understand

[charred merlin]

We speak English, and we understand english, or other languages.

AI speaks "token".

[tight spade]

and my back ground is building , so ai is very new to me

ahh i c

[charred merlin]

When we "talk" to ChatGPT or Suno, they automatically translate for us, so we never NEED to use this.

This just lets us see "the math" that Suno uses to "understand" our English, and turn it into "AI-readable language"

[tight spade]

hmmm

[charred merlin]

Unless you are doing research, you will never need this website

I have an example.

[tight spade]

thank god . i have headach

[tight spade]

go on

[charred merlin]

Very simple word , equals 1 token. Very simple for AI to understand what this token means

[tight spade]

i c ..

[charred merlin]

Add another letter, "hellLo", and sudden it doesn't understand so easily... It needs to break the word into "2 bites" in order to "understand it all"

[honest python]

Hi, please will a developer contact me. I’m blind and using a screenreader. I’m at my wits end with not being able to click the play button on clips, and not being able to use the model selector so can’t end songs easily. Please help. Am stuck on a song

[charred merlin]

@steady kelp ^

[tight spade]

@vestal rapids

[charred merlin]

Robert is not a developer

[tight spade]

no but he is blind

uses screen reader

[vestal rapids]

I am not blind, I am vision impared

[tight spade]

sry

[tight spade]

i meant nothing by it

[vestal rapids]

what can I help you with? I got pinged

[charred merlin]

Hi @honest python, I will send your post to the support lead, as I know a member of the team had reached out to you previously ❤️ 🙏

[tight spade]

@robert are you able to help a fellow with screen reader

[vestal rapids]

I can try, what's up?

[honest python]

Is there a way to consistently make songs play? Hitting enter to select, then using jaws cursor and left click doesn’t always work. And I can’t get the model selector to work.

I’m using Jaws

[tight spade]

@robert @honest python having issues chat with him until suno can get a hold of him

[honest python]

Yes the lead has not followed up

[vestal rapids]

I use Voiceover on a Mac, I don't use Jaws, I actually know little about it but I do know that it is difficult navigating the Suno site with a screen reader, I have needed a lot of in person help

[ionic wind]

Did they disable the ability to give feedback about sound quality? I was using those buttons a few hours ago, but they're gone now.

[honest python]

At this point if someone can log in as me to help me finish this damn song it would be great. And I’m happy to barter something

[charred merlin]

I just sent them a message now -- they have a lot of responsibilities as support lead, but I assure you that they will respond when they are able -- in the meanwhile, the most effective and direct way to receive assistance is by emailing them at [e-mail], as the formal support process through Discord is still in it's fledgling stages.

[honest python]

Maybe If someone logs in as me and changes to 3 it will stick on my end

[charred merlin]

Was this the opt-in feature with the smiley faces?

[ionic wind]

Yes

[charred merlin]

I know that was an alpha feature that not everyone had access to, it might be that testing phase has ended?

[ionic wind]

That was my first guess too, but I saw nobody talking about it

[rocky lintel]

Hey all. why audio upload doesnt work ( Network error) ?

[charred merlin]

As an alternative, is there a be-my-eyes-style app that someone could use?

I apologize I'm not comfortable logging in under anyone else's account, as I prefer to keep login-details private for everyone's safety ❤️

[charred merlin]

That indicates a general error with the internet, but it doesn't say if it's on your end or Suno's end. I'll try a quick upload to see if it's Suno-side.

[vestal rapids]

I log in with Discord, I have no idea how to log in as someone else

[rocky lintel]

its ok , i have to refresh the site

[charred merlin]

Suno-wise, everything seems okay.

There might be some things to look into on your end, internet-wise -- disconnect from a VPN if you're on one, pause any file-sharing programs, maybe try unplugging the modem/router for ~3 minutes and plugging back in

Keep in mind this is just community-sourced advice, [e-mail] is the go-to for the experts 😄 💪

[north sleet]

TeamViewer or Quick Assist which is built into Windows 10/11. Repost under correct thread.

[charred merlin]

Appreciated -- I'm sure the official support team uses tools similar to this 🙏

[north sleet]

I would help you @honest python with one of these assist tools but I am away from keyboard at moment.

[charred merlin]

As a general rule, we probably shouldn't ping visually-impaired users unless we have something actionable or intend to engage in conversation, as it's harder to quickly gauge context when you need to listen to every post 😅

[north sleet]

If they responded I would have said “let’s schedule a time this afternoon.”

I am just at lunch right now and helping from an iPhone is a no-go.

[honest python]

I’m free after 530 et amd happy to use remote assist

[leaden lark]

Hi tech chat

[north sleet]

That is 3:30 my time. I can do that.

[frigid cove]

when you have the lyrics you want but can't seem to generate the right music despite using the same listed stuff in style of music as a previous song you made

[tight spade]

do you have an example

[frigid cove]

Like... an example of the song I want it to sound like? Is that what you mean?

Sorry, I did not sleep well. 😅

[tight spade]

what tags have you tried and instruments genres.

[blazing quarry]

bro help i made a diss track about myself as a joke and at the end the ai started speaking gibberish and its rlly creepy 🤣 😭

[frigid cove]

I tried putting in the exact same ones as the previous song but the generation is different. The other song I made shows "fast paced progressive nu-metal rhythmic country western" with no commas/seperation between words.

[tight spade]

try adding commas between you type of music

[frigid cove]

Yeah, I tried, still comes out completely different.

Either way I'm honestly just going to keep experimenting. I have like 2,000 credits still.

[tight spade]

Sure! Here's how it looks with each genre bracketed:

[fast paced] [progressive] [nu-metal] [rhythmic] [country] [western]

i use gpts to do my dirty work

its had to help with out knowing a little more

this is what i put in lyric box instead of putting it into styles , i get more control i find '[note:[Genre: Rock, Indie Rock, Alternative, Classical, Orchestral, Symphony]]
[note:[Styles: Soaring, Epic, Reflective, Anthemic, Emotional, Driving, Symphonic, Cinematic]]

this is the lates one im working on

turned out really well.

[frigid cove]

so, uh, i actually came out with something good by adding the BPM XD

apparently wording fast paced, etc. didnt work good enough XD

[tight spade]

Nice !

[leaden lark]

I got something pretty fast with my one style prompt

[tight spade]

share

[leaden lark]

It’s all in Japanese 1 sec

[tight spade]

np

[leaden lark]

https://suno.com/song/1e095941-cd2f-4b80-abc1-b70aff459aa9

[tight spade]

holy molly .. my braiin zappped

so fast

[leaden lark]

The power of that style prompt

Does that to the songs

[tight spade]

wich one exactly ?[Cyber Japanese-style EDM]: [High-speed Taiko Drums] + [Shinobue Synth] + [Shamisen Bass], [Koto Glitch Effects], [Shakuhachi Drop], [Kabuki Scream] + [Enka Rap], [Neon Ukiyo-e Sound], [Samurai Techno], [Oiran Trance], [Dubstep], [Trap]

song is amazing

[leaden lark]

Not sure, but I love it

[tight spade]

woooi have a tech question , y two gens ata time y not one?

[charred merlin]

1 sec, recording quick video clip to help describe.

AI is trained to recognize patterns, but doesn't actually KNOW what music sounds like -- it doesn't understand what "rock music" sounds like, but it has studied enough songs that called themselves "rock" for it to recognize similarities in how the guitar "looks" when it's viewed as data like the sampling dots in the video. --- to be continued after pizza

I'm stepping out to grab a pizza right now, but I'll write up the rest of the description in #1270928841614823546 when I get back ❤️

[tight spade]

enjoy

[leaden lark]

That’s true, ai has no clue what the music actually sounds like, it can only go off of what it knows from its training data clues from that data

[steady kelp]

Hi! I'm that lead @charred merlin mentioned, but unfortunately, we don't have a ton of great feedback regarding accessibility for vision impaired users at the moment. Our primary support channel is email, [e-mail], and we'd be happy to hear out any suggestions for how we can improve accessibility. The team does not utilize remote desktop applications just yet, however, so in-person assistance may be the easiest at the moment. We can work with our engineering team to provide some insight if you still need help- feel free to email the team and we'll do our best!

[tight spade]

@steady kelphey mike iv heard people with vision issues having a hard time with navigating all the updates, the more that's on the site the harder for them, could it be possible to just a stripped down basic page with larger buttons and min text just for the vision impaired? less is more i hypothesize

[steady kelp]

So, personal opinion only here (not that of the company, I havent consulted devs on this), but I'd imagine a whole separate skin for this might be less valuable and potentially more time consuming to accomplish than improving the current state of readability with certain core UI features

[tight spade]

thanks for the reply, understanable .

[steady kelp]

I think to your point, there's been a lot added recently. so taking the time to go back and reevaluate what can be made more accessible is really important. Having an easier solution for the end user is certainly a viable answer, too.

[north sleet]

Hey, I set a schedule here shortly to use a remote assist with vocativecomma to try to help solve the issue. I can share my findings on what is tripping up the screen reader.

[quick agate]

dude, found an ai service that can extract instrument stems
vocals, drums, guitar, etc

[tight spade]

is it free'

[quick agate]

don't think so, im testing it out now, but its a one-time fee
https://www.lalal.ai/pricing/?_gl=1*binu12*_up*MQ..&gclid=CjwKCAjw_Na1BhAlEiwAM-dm7HQoSFiT0f-GD8AFnFG94irzkhPL8_RfRu-QChrDIswG_us9OCuw_xoC4IIQAvD_BwE

[tight spade]

i use moise plus its 6.50 cad a month for unlimited stems with same stuff and more

https://moises.ai/ iv tried lalal ai its exspensive'

[quick agate]

its a one time fee, not monthly

[tight spade]

check it out if you like lots a cool features and extras . 6.50$ a month is cheaper iv done a 150 stem wtih it already

[vestal rapids]

Stem splitters, no matter who makes them, are imperfect. The only way to get stems is at the production source

[tight spade]

how much is that ?

[quick agate]

all I care about is which one is cleaner and more accurate

[cunning siren]

I wonder why the pieces are not generated thru stems right from the beginning. Isn’t that easier than creating complex waves?

[quick agate]

no, its not

you have to remember that the goal is to create a product that is consumable ready
so it uses a fuzzy generation on the expected output, building stems requires foresight into the final composition.
its easier for a machine to go backwards, think of it as how image generation works.
ai wouldn't do each brush stroke or critic itself without a goal of the final image

its far better to have a fruityloop ai plugin that can sample songs and generate instruments based on guesses and filters

[tight spade]

can o worms ... over my head

[quick agate]

basically, can you predict the future with a noise generator?
answer is no.

[lunar anvil]

That is why generative AI is a good blueprinting tool help with the creative process.

[cunning siren]

Frankly, for me there is no real difference. It‘s making things up in this or the other way with one difference: on dedicated „tracks“. So instead of putting his „creativity“ in one pot it may very well spread it around. What is true is the mixing process would take longer but I believe it’s for quality’s sake 😊

[quick agate]

ok, so lets make it simple.
the ai only has 1 wave form to work with.
multiple instruments means multiple tracks, and multiple wave forms to join.

its like 2D vs 3D

you can't get an oil canvas to do gpu shaders

[lunar anvil]

Getting into Flatland to describe stem separation?

[quick agate]

I loath flatland so much but yea

[cunning siren]

I keep optimistic about that - sooner or later this will be possible. Think about the possibilities and the raise in quality 😉

[quick agate]

yes, its 100% possible, its just a secondary post process. like a puzzle with the front of the box art to reference

[lunar anvil]

Wish Suno would release the API or docs on how to shape the AI a bit better. Taking a stab in the dark is the most frustrating part.

[quick agate]

100% agree, would like to see more of a bite wheel and samples to reference, like a virtual project rather than a ringtone generator

[cunning siren]

For starters the seed would be nice

[leaden lark]

Has anybody tried to do instrumental call and response? I’ve tried many times to little success

[charred merlin]

@tight spade https://www.bandlab.com/splitter is free, and amazing 💪

[charred merlin]

This would be way more complicated, tbh, and would likely need to be designed that way to begin with -- not something that "suno should do, but isn't", it would likely be a whole new project, which is still something they might consider?

Are you requesting technical feedback on anything, looking to demonstrate a particular method or is there a section of your songs that highlights an example of a method you learned from another user?

Or are these songs that you're proud of, and wanted to share them with a wider audience?

[midnight swan]

Ah crap on a biscuit, stupid phone my bad

I was wondering why it didn’t show up in the showcase channel just now haha

[charred merlin]

No worries, I was just checking -- it's a quiet night so it's not bugging anyone 😄

[cunning siren]

If they want to get their foot into the DAW biz, they might consider that

[midnight swan]

All fixed, sport about that again

[charred merlin]

The devs regularly pop their heads into #💡┃suggestions-web if you wanted to share that suggestion -- they've expressed interest in knowing what people's workflows are. 💪

[charred merlin]

Oh -- sorry dude, I didn't mean to make you delete the post! 😅

[midnight swan]

I moved it to the correct channel, so deleted it from here

[midnight swan]

My god I can’t type today….

[charred merlin]

Cellphone keyboards have a mind of their own -- and they're all designed to be at least 30% evil.

[charred merlin]

😅 Still, I apologize if I came across more as "guard dog" than "elderly walmart greeter" 😛

[midnight swan]

Oh I def know that, I asked my wife yesterday if we had Prilosec still and didn’t notice it changed it to Prius Sex until she called laughing because the car read my text while she was driving with coworkers in the car

[midnight swan]

Oh no don’t worry not what I thought at all, just didn’t want to have the same thing posted in multiple channels and have woodhouse yell at me

[leaden lark]

Hey harber, can I dm some lyrics I just cooked up even though I can’t make any songs rn?

[rose jasper]

It's ok to have creative gaps!

[leaden lark]

I can t make any since I have 5 credits left lol

[rose jasper]

Oooh, hahaha! Sorry!

Shame we cannot send credits to people

Oh! New idea!

Award tracks with credits!!

[leaden lark]

Possible, but that could get abused with botting

[rose jasper]

Yeah, well, I'm sure the devs can fix that

But I have 8500 credits, if you want to use mine 🙂

[leaden lark]

For sure

[rose jasper]

You can add me on Discord if you wish

[leaden lark]

I’m fine, my refresh happens in a little over a week

Plus next week is pay week so it’s all good

[rose jasper]

How is your profile doing atm, still trending btw?

[leaden lark]

Nothings trending rn

But otherwise not bad

[rose jasper]

We are kind o the same, haha But I have WAY less fans

It's so hard how the algorithm works, and getting noticed, but it does not matter. We just keep on making good things

I'll toss you a follow though, would love if you do the same ❤️

[leaden lark]

Done

I can’t believe my churro song trended twice lmao

[rose jasper]

One of my most favourite tracks is about going to the toilet, so I get what you mean hahah

[leaden lark]

This one trended a while and I made it as a joke https://suno.com/song/855bed2c-ab78-4bda-92fb-9c62261720c4

[rose jasper]

IKR! haha!

If I can offer you one piece of advice: K-POP

It lands plays

345 plays now:
https://suno.com/song/50357e8c-e332-44ca-9c94-ffe8a55c4c64

[leaden lark]

I’m not too huge into k-pop

[rose jasper]

And I have been using this one now, learning by what I have learned from proper and accurate prompting, by the asistance from the lovely @charred merlin @wet minnow @untold briar They've been of huge help in my progression with thing

[leaden lark]

I’m more into edm genres, and genres like j-rock, but glad you’re improving!

[rose jasper]

Yeah, but everyone else does.

[rose jasper]

I've listened to a few, you prefer slower and ballads?

[lunar anvil]

I just listened to your Am I track earlier today @rose jasper . I don't remember how I found it.

[leaden lark]

That’s just what seems to somewhat pop off for me recently, there’s a reason people know me as the chill person

[rose jasper]

Aww, really?! I'm so proud of Drew Everlong, I will actually present that entire album and have developed with me singing ❤️

[rose jasper]

https://suno.com/playlist/658cf5cb-2612-4f86-89fc-7f06b61862f4

Here's the playlist, many good songs, all written by me, word by word. And directed by me. I have four songs on all musical platforms, look for: Drew Everlong

[rose jasper]

I love making Pop songs, take a listen to Drew and let me know what you think, I excel in Pop songs (I'm Swedish)

[leaden lark]

This is one of my most recent ones that trended for a little https://suno.com/song/6a151181-a8ec-41ea-b6de-0df98959b469

It’s atmospheric dream pop

[rose jasper]

Did you write this, yourself?

[leaden lark]

No, I had some help

[rose jasper]

A great recommendation from my persective, if you are at a break, and you don't know how to progress, everyone has a few request that we can all use from the latest version of ChatGPT, and sometimes I ask it how to make a song better.

Also, I keep on mentioning this, but it's so helpful:
https://www.suno.wiki/wp-content/uploads/2024/04/PROMPTS-FOR-AI-DANCE-MUSIC-🙏mista.lewys_.pdf

[lunar anvil]

I use that PDF as a reference all the time.

Some of the tags work some don't. Sometimes they work and the AI vocalizes the tag which is very annoying.

[rose jasper]

omg, it's so good, just to find new ways of working with Suno.

[rose jasper]

I know, but what I've discovered it does do this way less now with the latest "hot patch" of 3.5

[charred merlin]

Yup! 😄 You can always DM 🙂

Sorry for the delay, just happened to scroll up & saw my name 💜

[rose jasper]

I really like this hook: "She’s got that alchemy, turning heads with a flair,
A symphony of fashion, a vision so rare."
@leaden lark It's so catchy

[charred merlin]

You rock, @rose jasper 🤘

Always glad to see your name pop up in chat, whether it's sharing feedback on something new you're trying, helping others in turn, or swinging by to add a kind word.

Thanks for letting me play a small role in your Sunorchestra 🙂

[rose jasper]

Don't ever, ever sell yourself short. You are a part of my life at this moment, and I have no plans of letting go ❤️

[quick agate]

any AI that could do monster sounds?

I have some monster calls from audio files, but I kind of want to make 20 varients of those audio files for the other 20 creatures

[rose jasper]

I am not sure if this is possible, as there is no sound database to my understanding @charred merlin

[vestal rapids]

@charred merlin I just posted something into #🗣️┃chill-chat which demonstrates Aiva

[rose jasper]

Can you link his "specific" post please? "Copy Message Link"?

[lunar anvil]

Do you clean up your lyrics after generation to remove the prompt details and only show the lyrics?

[rose jasper]

I do

In my understanding, everyone should learn to prompt with their own experience.

I save all of my prompts locally

[lunar anvil]

I haven't got around to clean any mine up yet. I need to copy my prompts for safe keeping then clean.

[rose jasper]

And if I make non-English songs, I have a tendency now to translate them over to English

[vestal rapids]

#🗣️┃chill-chat message

[rose jasper]

You should! Really! This is very important, we don't know when people "start to learn" or "steal your prompts"

[lunar anvil]

Very true. Maybe I will do that this weekend.

[rose jasper]

Always, only include the basic, never show anyone how your prompts are made, if you expect to have this "your own"

Make people come to YOU

My prompts are incredibly inctricate and complex, but I have learned these from spending countless of hours.

[lunar anvil]

I used [Modulation] tag but left the front bracket off by accident and didn't notice until about the 10th generation. Even after fixing it the AI will randomly sing Modulation.

[rose jasper]

Let me share you one song:
https://suno.com/song/3796766a-ac29-40fe-8bfb-8453b4490c9a

[lunar anvil]

lol

[rose jasper]

See how many things I am prompting Suno?

It's very in-depth

[graceful condor]

Hello!, Did anyone deal with an unwanted echo on Suno's songs?

[lunar anvil]

Looks pretty clean to me.

[rose jasper]

Include [Isolate Instrumentals] before the start of your vocal performance, you can also include this whenever there are echoes being made by Suno.

[rose jasper]

You mean: a6dd8d67-dc42-4091-9752-2eecf43b70b1?
#📚┃suno-school message

[lunar anvil]

Didn't see that. @flint haven removed the link.

[rose jasper]

Aaah ok!

[lunar anvil]

I see it now by going directly to it.

[tight spade]

@rose jasper yuh man, dont fret , its clean . compared to some . its what the end result matters

[graceful condor]

It's instrumental and the 'main vocal' is a guitar, I believe it's applied to the guitar somehow. Do you think that would work?

[rose jasper]

So that's how I make my tracks, I spend around 2-3 hours when I am making one song, and then I remove my prompts

[rose jasper]

What are your "Style of Music"?

[graceful condor]

Stoner rock

Minute 2:22 for reference: https://suno.com/song/b8f2c44d-9b29-40d1-8080-4fe0c6c167f2

[charred merlin]

I'm not familiar with too many different AI audio services, though while I haven't seen metatags, there's some absolutely horrifying noises you can make with Suno. 👀

If someone were to clip those noises out separately in a DAW, they'd have a bespoke pig-demon soundboard within seconds. (https://suno.com/song/197cffc1-da6e-4fb6-afe7-362b62f0c84b)

It seems that instead of JUST music, anything that one could reasonably expect to be commercially-available on longform audio seems to be reasonably promptable -- SixStringToker made a really neat weather-based generation, and I tried to metatag-prompt some lightning, it's in the pool party > #1256301881966198805

[rose jasper]

I see what you mean, these echoes are with the instrumentals! Try to work more with establishing with reverbartion and cleaner sounds like [Studio Quality] [Clear Audio] [Less Soundscape]

[graceful condor]

Ok, thx! I'll try those

[rose jasper]

Yeah, since now, you are more directying it to act in a more "open way" like... "I want to make an instrumental track, but give you full freedom with quality" So you just need to be more directive with it.

[rose jasper]

Truly a horrible listen hahaha

[north sleet]

This one of my instrumentals that is pretty heavy on the prompting. https://suno.com/song/0e9737f5-5166-4b23-bce4-92074e3b46c3

[charred merlin]

Checking it out now robert ❤️ Thanks for sharing, I know that it's clearly an important service to you, as I recall saying that you volunteer(ed?) for their Discord as well 💪

It's a snowball effect once you start to learn about one service, it makes it a lot easier to understand and explore the offerings from their competitors.

10am on your end, right? I hope you've sufficiently thawed out from last night 😂

[lunar anvil]

I am working on another one, but haven't been able to nail it down 100% yet.

[charred merlin]

I put it on for @delicate idol with a cheeky-grin, and she just instantly raised her voice to the "I'm not joking around" level, and simply said "Nope."

I turned it off instantly.

[vestal rapids]

yes, 10:10am here now I have been sitting in my recliner with the heater going 🙂

[rose jasper]

Please keep in mind to use [hooks] when it comes to these tracks, as most EDM and IDM songs are meant to be made with some breaks, sometimes they are not enough to have downwards progressions and upwards progression.

The song would be much easier to listen to if it had breaks, such as hooks, and I'm sure you would love the future result from it!

Otherwise, I find it as a chaotic Progressive Track, but it's geting somehwere.

[vestal rapids]

and the drop!

[tight spade]

@north sleet umm hi, @charred merlin should i show her my work flow

[rose jasper]

Hahhaa, tormenting the wife I see

[tight spade]

lol

[rose jasper]

Another note, you have the possibility of using any genre you want, I would personally throw in some more random lyrics and even do a genre shift in between.

Considering the song it already 8 minutes long, this is more of a festival stage presence sort of song, and with two large silent breaks in between, this is more SUNO working in it's weird ways as well.

[charred merlin]

If they're requesting help and you feel like you may have some expertise or insight on the matter, it's a nice thing to do 🙂

If you find that you end up sharing the same ideas with multiple people, it can save you some time to create a post in #1252267444773326848 to gather your thoughts -- plus that way you can slowly work on the posts, add to them, and make them look nicer over time! 💪

[charred merlin]

I'm living in both worlds -- I like leaving the original videos up so people can see the image changes to pick up on extension points, but I love having the ability to "mask" some things to preserve a bit of mystique 😶‍🌫️

[north sleet]

It was mostly an experiment of trying to steer Suno.

[charred merlin]

https://tenor.com/view/steering-wheel-broken-freak-out-gif-10741306

[north sleet]

Also, it is 8 minutes as it was purpose built to be long. It would have been longer, but it was exhausting building up 30-46 second extensions so I just merged and called it good enough.

[full field]

i cant get suno to give clear voices. it sounds horrible. from 1-10 its like a 2-3 for clearness

https://suno.com/song/168f0bab-7e72-4f33-af58-6c8abd8e7434

wth

[tight spade]

try any of these :
[narration]: The act of telling a story or giving an account of events.
[dialogue]: A conversation between two or more characters.
[monologue]: An extended speech by one character.
[voice-over]: A voice that is not part of the narrative and provides commentary or context.
[intonation]: The rise and fall in the pitch of the voice during speech.
[diction]: The clarity and distinctness of pronunciation in speech or singing.
[accent]: A distinctive way of pronouncing a language, often associated with a particular region or country.
[inflection]: A change in the pitch or tone of the voice.
[articulation]: The clear and precise pronunciation of words.
[projection]: The strength and clarity of the voice, especially in a large space.
[pause]: A temporary stop in speech for effect.
[emphasis]: Special importance or significance given to a word or phrase.
[cadence]: The rhythmic flow of a sequence of sounds or words.
[rhythm]: The pattern of sounds and silences in speech or music.
[tone]: The quality or character of the sound.
[pitch]: The perceived frequency of a sound, determining its highness or lowness.
[pace]: The speed at which someone speaks or reads.
[fluency]: The ability to speak or write smoothly and effortlessly.
[delivery]: The manner in which a speech or piece of music is presented.
[expression]: The conveying of emotion or meaning through speech or music.
[clarity]: The quality of being easy to understand.
[volume]: The loudness or softness of the voice.
[modulation]: The variation of pitch in the voice.
[resonance]: The richness and fullness of the voice produced by the reverberation of sound.

or combination

[charred merlin]

...prompting text-to-speech from a model trained specifically on music is amazing in-and-of itself, but if you're in the frame of mind that it's a terrible generation, instead of framing it as a comparable comparison to the average Suno-generated standup, it's going to be hard to give an answer, or actionable feedback that properly expresses the gap between "what you're expecting" and "what Suno is largely capable of at the moment".

It seems like you might be more interested in Suno's now open-sourced text-to-speech model, Bark (https://github.com/suno-ai/bark).

Alternatively, you can flag these generations with ... > Report > Bugs > Poor Audio.

[tight spade]

try or fail these some work some don't experiment

[tight spade]

would these be appropriate here, ? there to do with music notes delete if bothersome

[charred merlin]

@tight spade, even if this may be subjective to each listener (some people might find different keys to evoke different feelings), I know some people have a difficult time thinking of new ways to describe music -- they might listen to a powerful rock song, but not know how to describe it.

Looking at lists like this could help people think of different "thematic descriptors" to add to their metatags 💪

"uneasy" sounds very evocative, I'm going to try adding "uneasy piano" to a style prompt and see if I can imagine what influence "uneasy" might have had on the song, or if it sounds like a "typical piano song"

https://suno.com/playlist/ab237205-f8ac-4318-9bb2-d4177b0c16a3

I mention it in the playlist description, but for visibility;

Perhaps confirmation bias, but the v3 songs seem to have hesitance?

[tight spade]

may i suggest some thing along this to test i am terriable at the science thing. [note:[genre: instrumental ]]
[note:[ tonic: EB major] <<<< or what ever mood your going for

of you can reference this small test by changing the tonic. https://suno.com/playlist/d13399df-b5e5-42e7-8247-6ac042b76657 i dunno peer review i guess

[leaden lark]

Sent a dm with the lyrics for something once I get the credits

[full field]

@charred merlinfound the problem. its 3.5

3 does fine

but 3 doesnt have enough character space

1250 isnt really that many characters

[charred merlin]

I guarantee you’re already a scientist in real life without knowing it!

Anytime you try a new ingredient in a recipe, or if you add twice as much of something, that’s an experiment — the “science” part of it is how you think back on your results — when you added twice as much salt, beyond “good or bad”, did it taste better or worse than last time?

As long as you pay attention to the results of your experiment, and try to use it to help you understand how to do things next time, you’re a scientist! 🧑‍🔬

I know that I tend to go “overboard” when trying to understand Suno, and like to break the problems down into very small problems, but not everyone needs to experiment the same way I do — in fact, it’s better that people experiment in their own ways, that way we can “look at a problem from all angles”. 🔬

Thanks for putting on a labcoat and being a music-scientist with me 😄

🥼🎧🤘

[hollow hare]

Another important thing about science is reproducibility.

I thought I had a good way to duet but subsequent Reuse Prompts disproved that (though it had quite a number of partial successes). Test series: Harmony in Tune at https://suno.com/playlist/11ff3827-bdcf-4ef5-bd04-da8647c6cdb6

Surprisingly, another test series I did (with more paragraph gender voice rather than line-by-line that HiT had), the Big Mac Dreamin', actually had pretty great reproducibility.

That is how I found out that most genre changes do not affect the duet prompts much but adding different gendered voices in the style prompt generally failed (likely because 'male' has a very low impact/priority)

[charred merlin]

Yeah, the character limit is designed to try to limit how much Suno needs to “think about” all at once to prevent it from hallucinating more frequently.

Think about that big ridiculous article in the news last year about the journalist who spent 2 hours chatting with Microsoft Bing chat bot with no safety guiderails (back when the public-facing bot just terminated the conversation after 5 exchanges to prevent hallucinations) — he TOTALLY overloaded it, and it got confused about being asked if it loved the user, and asked him to leave his wife (user loves me, I love user, user is married, possible solutions? ) — the limits are there to stop things from going bonkers, but it can still be annoying to work around.

[charred merlin]

Reproducibility
I swear I’m gonna get that tattooed on me somewhere. Like the nerdiest barbed-wire style bicep tattoo in the world.

[full field]

sure but why didnt they increase it to 3000 like 3.5 is

[hollow hare]

Ugh, thought I was spelling that wrong, thanks

[charred merlin]

…didn’t they double it?

[full field]

nope

still 1250 on 3

[charred merlin]

Yeah, they can’t improve the old model — that’s just the old model. The new model v3.5 has 2500, no?

[full field]

no

[hollow hare]

Different model have different training and parameters. They would have to change quite a lot of background stuff if they wanted to upgrade the character limit.
(Not to mention it is useless to have more in 3.0 because of the low timing)

[full field]

3000

[charred merlin]

Legit didn’t see if you spelled it differently, I’m rolling into bed, I was just giving support because I whole heartedly agree 🙂

[charred merlin]

So they did better than double it — I understand you’re saying you want everything else about v3 to stay the same, but just to have a bigger context memory — suffice to say, only new models “get better”, everything except for the absolute newest model is very likely to stay as-is unless otherwise noted.

[full field]

i mean u can do alot in 2min

[lost flame]

@charred merlin I finally got it, getting male and female in the same song!

https://suno.com/song/fdc0f870-380e-48ef-9a1b-1ba00737c733

it changes from female to male to female to male, in a perfect order no overlapping and no weird bugs

[leaden lark]

I got something similar in one of my songs and I wasn’t even intending for it to happen

[lost flame]

I was trying to find a vibe to it right, no style of music as prompt just hints and it finally hit, after like 2-3 days hahah

[lost flame]

also could you send me your song please :) Im really curious

[leaden lark]

Dm? Or in here?

[lost flame]

technically its technical so would make sense here , but its up to you :)

[leaden lark]

1 moment on mobile so it’ll be a sec

This is one example where it flips vocals https://suno.com/song/0ea1f38e-1d31-4bfc-8eb1-fdec484ba5a6

I initially wanted only female videos

[lost flame]

I see the emotional female vocals ahhaha, AND NOW its a dude ahhahaha

wow

that's funny

[leaden lark]

It honestly makes it better imo

[lost flame]

yeah! but still funny that it just gives up on your request haah

its like, no you want it like that not like that

[leaden lark]

Lmao

1 sec got another where for a singular pre chorus it’s a legit duet

[leaden lark]

True, this one got a real duet around 1:18 for a singular pre chorus

https://suno.com/song/0ab36074-83de-4de0-8bb9-71bcdb78df7e

This was testing to see if duets were possible

[lost flame]

yeah damn that's really mixed, interesting !

[leaden lark]

My face was so shocked when I heard it the first time where I kept replaying it

[north sleet]

I was shooting for a Billie Holiday sound when I got this perfect duet. https://suno.com/song/d3d5ff6d-36ff-4ea4-aea3-b49bafa90128

[lost flame]

now the real challenge, keeping it so that one makes the verses and the other makes the chorus and at the end both sing to each other like one pair of words each, for example bring me to life - Evanescence

[lost flame]

gros c tight, mixed really well wow

[cerulean kestrel]

Sorry bro, I straight up went to bed lol, it was 2am for me. Kmac is the goat tho

[north sleet]

I have another, “Shadows of Doubt,” that flips back and forth between normal voice and screaming voice.

[lost flame]

OMG I wanna see that foshoooo

[north sleet]

My “Dreamweaver’s Journey” does this. It sounds like both voices are singing the chorus together. And “Lost in Motion” flips back and forth in the same line in the bridge.

[lost flame]

taking a listen right now

[lost flame]

when you wrote " barbershop " what were you looking for?

ohh ok ok I see

[north sleet]

A cappella or harmony

Sorry, was trying to keep it the same thread.

[lost flame]

cuz all my barbershop where I live near only play R&B hahhah

[lost flame]

no worries it aint reddit x)

[north sleet]

This is a wonderful song. I wish I knew Russian to truly get to experience it.

[lost flame]

mostly the same vibe as the english parts ;) just imagine what it could say :)

[north sleet]

It flows so well. Reminds me of the track Loading 99 by PERS MANTRUM and The Anix in how it mixes English and Russian.

[obsidian atlas]

Has anyone tried extracting just the melody and changing the lyrics to something else? It's not that weird if it's very slightly different, right?

[lost flame]

I'm a big metalcore / deathcore fan, on my calf I have bring me the horizon and slaughter to prevail, and slaughter to prevail ( alex ) uses a lot of mixing between russian and english and I love the vibe it shoots off ! take his latest song for example ( " Kid Of Darkness " ( Which is literally the tattoo I have on my leg which is pretty funny ) ) but he mixes english with russian and I f*cking love it and is what inspires me with most of my recent songs

dont know if I can but here the link of the song Im refering to " https://www.youtube.com/watch?v=dNJtJrbFeWw "

I want to explore spanish with french soon, but cant find any inspiration

[lost flame]

you'll probably get a repression sound on your track since its muting vocals but trying to push instruments in the background

[north sleet]

I started French, German and Russian earlier this year but I had to pause as work got crazy busy. I was using LingQ to learn the words.

[lost flame]

if you need help in french I'm here! can't help you with russian since I still make mistakes, but french is my native language ;)

[obsidian atlas]

Then... just re create is recommended?

[lost flame]

you'll probably never end up with the same track, so if you got music talent maybe just remake the beat on flstudio ( since its free ) and use suno feature of importing instruments and rewrite lyrics

could be rough but a challenge is always good to improve skills

the more time you struggle on something and figure it out on yourself the better you'll get

[obsidian atlas]

FL studio???🫨🫨 I never heard about it... oh
Thanks bro. I will try

[lost flame]

dont know if you're sarcastic or not, but if not well no worries man ! <3

[obsidian atlas]

Or.. just i should sending this time nice getcha! 😭😭

[north sleet]

You could extract the melody then match with midi in a DAW to clean the stem then bounce to audio track then feed it back into Suno. Or use Ace Studio with the licensed voice models to have a different voice.

[obsidian atlas]

No sarcastic. Nono

[north sleet]

That’s cool. French is my dad’s native language but the nuns beat it out him when his family moved to California.

[lost flame]

@obsidian atlas 재미있게 지내고 여행 잘 다녀와! ( hopefully it's not insulting you ahhah )

[lost flame]

I feel the nuns AHHAHA

I grew up with english so everything I watch or listen is always in english and well when I arrived in french high school I couldnt understand anything my french teacher was speaking about since I didnt learn half the terminology of grammer haha, still at the end of high school I failed my ministry exam with 55% and they made me pass ahha

[north sleet]

BTW, FL Studio fruity version doesn’t allow import of audio tracks. Only Producer and above bundles allow audio import unless you have an ear to match.

[north sleet]

The one semester of French in high school didn’t sway me to go farther with any language.

At least at the time.

[lost flame]

I feel ya it's horrible HAHA and now I'm losing my english since the quebecois in me has been awoken hahahah

[north sleet]

lol quebecois, eh? I have a bit of extended family up yonder way that I have never met and probably never will.

[lost flame]

you should, just to visit, its a beautiful place!

[north sleet]

Might be a good idea to burn some of my 600 hours of PTO I have saved up.

Have a nice night/day. I’m off to sleep.

[night jewel]

SOrry to ask this question again best tool to seprate vocal and bgm ?

[tight spade]

whats bgm?

@night jewel you can use the stem separator in the menu separate vocals and music

[foggy wing]

Great work! What does the [suno.ai: dolby atmos] tags do?

[charred merlin]

https://suno.com/playlist/c6fb3c02-2dd2-455d-93bb-050c7fa8f9fd

Jon Fly and I worked on a prompt experiment playlist, and from my observations, I haven't noticed any difference -- I asked a few others with better audio setups to compare, and there wasn't a discernable difference in audio quality.
That being said, at worst it's a "thematic decorator" like any prompt you use -- I think Jon fly was under the assumption that it added to a certain "movie soundtrack" sort of influence.

As always, more experiment playlists in the pool party > #1256301881966198805

[charred merlin]

"background music"

[tight spade]

ahh i c ..

cool

24hrs up again i have some cool stuff to reveal or i might need help to reveal

[foggy wing]

I have tried with similar lack of results with tags like [high quality], [lossless] etc.

[charred merlin]

If you're a paid Suno member, you can use the "Get Stems" option, but I'm a big fan of all of Bandlab's free web-suite of tools ❤️ 🙏 💪
#📚┃suno-school message

[charred merlin]

Yeah, it's like adding "masterpiece", "best quality" to an image-generation prompt, which admittedly I still do depending how the wind blows.

That said, the tags we tried were common enough in music that they wouldn't have been too obscure, but targeted enough that someone with a better-than-average audio setup could quickly determine.

[foggy wing]

The difference is that is has great effect when done in image generation. (high detail):2.0, (best quality):2.0 and inverted equals for negative prompts are standard when I create visuals.

[charred merlin]

😛 I realize it's not a perfect analogy -- I tried add the caveat that I still do that for image-gen 😅

[hollow sapphire]

👏 I am quite satisfied with this result; it seems that the AI is finally becoming aware that there are more precise prompts when they are integrated into the texts! What do you think? https://suno.com/song/cb9b71a1-bcf4-4906-ab7c-d665e3f94f4e

[foggy wing]

Yes, it can be done with great effect. However, it is a hit or miss. This took over twenty tries before Suno actually put the right stuff where it was supposed to, but it still completely ignored the notes for male/female vocalist. The chords probably do not match the output either, but it just feels better to put them there.
https://suno.com/song/6579d35e-69e6-4100-a838-4a53d6ab395e

[golden abyss]

Not an extend, many prompts in the style section, yet i get a 19 second song.... now that's a waste of credits!!

https://suno.com/song/456ff5ab-b3cb-428f-a149-5dbd597e7445

[turbid patrol]

I'm trying to make a remix of a public domain song Entertainer by Scott Joplin from 1902 using Suno. When I upload fragment I get error "Uploaded audio matches existing work of art.". Are we not allowed to use Public Domain audio?

[golden abyss]

and since he passed in 1918, his work is DEFO in the public domain, worldwide.

[hollow sapphire]

It's still a lot better than at the beginning, even at the start of version 3.5. I remember that the AI used to produce either really good stuff right away or complete nonsense, just doing whatever it wanted! Now, it feels like there's more consistency^^ That said, I've managed to get some pretty decent results with "duet voices" or "dual voices," or even by specifying "male voice" and "female voice" in the header, and then using "voice 1" and "voice 2" in the lyrics. I’m not sure if this has happened elsewhere?

[charred merlin]

@golden abyss it's not a matter of whether or not it's public domain -- Suno only care about whether YOU are the ONLY PERSON that has the rights to use your audio upload.

If it's available to more than 1 person, multiple people upload it and use similar prompting methods, Suno will likely output similar songs.

It's not a matter of "whether it's legally permissible to use" in other projects -- Suno only cares about you being the only person in the world who is legally allowed to use your content.

[turbid patrol]

Got it 😦

[charred merlin]

Here's a fun workaround -- the song IS public domain, so there's nobody you need to ask permission from in order to make your own version 😄 💪

[golden abyss]

hehehehe like i did with scotland the brave 😛

[turbid patrol]

lol

[charred merlin]

I'm not a legal expert, and wouldn't want to misrepresent the situation, but you could look into what degrees of changes need to be made to be considered a legally distinct work, then while you still need to acknowledge the sampling credits when publishing the song, you'd be the owner of that -- even if it came from public domain, you'd have created a legally-differentiated version that people would need to ask YOUR permission to use 🙂

[golden abyss]

hummm not sure if you could claim the rights to a alternate version, maybe to the music part of it, but the lyrics are now public, i'd have to dig deeper into coopyright laws 😄

but the entertainer is music only, so go crazy 😛

[charred merlin]

As in general legal terms, or for use in Suno?

[hollow sapphire]

🇨🇵 The rules vary greatly from country to country. For example, in France, no one will say anything if you use a very small portion of text or even a tiny piece of music, like a chord or a very short and non-repeating melody. That said, no one is entirely safe from a plagiarism lawsuit, because since the dawn of time, stories and sounds have all started to resemble each other at some point! That’s life^^

[charred merlin]

You're absolutely correct, region plays a huge role in copyright legalities -- I just wanted to highlight that simply because it's legally-permissible to use a song elsewhere does not mean that it is legally-permissible for use in Suno at the discretion of the developers.

[hollow sapphire]

I approve, that's right^^ 🙏

[charred merlin]

It's SUCH a tricky field -- I know that existing artists need to be very careful when listing their samples, tracks used, or collaborative partners when publishing songs, so I wouldn't be surprised if some of the horror stories I've heard about DistroKid dropping partners or YT accounts getting bricked are from examples of people rushing into the use of these tools, using methods that aren't legally-sound at the moment.

I'm glad we can have a nuanced discussion about this, because I worry that trying to explain the "line in the sand" is viewed as a strawman argument equivalent to "you're not allowed because it's not fair to use other people's music" -- I'm so interested in figuring out these legal distinctions, but because it's SO EASY to create these days without needing to put much forethought into the process, people can easily forget that Suno is a private company that is well within it's rights to legal protect themselves or their users by moderating the content they accept.

Bon Matin-Canadien, @hollow sapphire ☕ 🙂

[hollow sapphire]

I have fully understood Suno's rules, but it is true that these rules are specifically designed to protect the company, which is entirely logical—why expose yourself to lawsuits when you can avoid them! Personally, I don't see any constraint on creativity. However, if I mentioned the different statuses of intellectual property around the world, it's simply because these are facts that some may not fully understand. I cited France as an example because those are the rules I know, having been a "lyricist" in my time. But it's important to remember that these rules do exist!👍

[hollow sapphire]

Donc, tu parles Français? Et chez toi, c'est le matin? Moi, il est 15:40 ^^ 🇫🇷

[cunning mulch]

how goes the experimenting with Extensions and Displayed Lyrics?

[charred merlin]

💪 Your insight on the matter was legitimately helpful -- because it's such a complex matter, some prefer to see the issue solely as black-or-white, so it's refreshing to have nuanced discussion around the matter -- like you mentioned, many people have mentioned that they don't think to ask these questions, but they really enjoy being able to "take part" in these discussions.

[charred merlin]

Techniquement, mes notes ne sont pas mauvaise, mais mon Francaise a 20 ans 😅

I know "just enough to embarrass myself" 😂

[charred merlin]

Hey @cunning mulch, hope your day's going well!

The edited-lyric test is still upcoming, though the revised testing methodologies you shared in DM will make it a lot easier and take some of the guesswork out of the experimentation phase -- I'm still planning on giving you first notice when it's ready to be shared ❤️

As domineering of a presence as it's taken in my life, Suno is still a hobby, so I'm trying to balance professional & social responsibilities at the same time 💪

[hollow sapphire]

Ha ha ha !!! Yes, I admit, I love dissecting laws, not to flout them, but to better understand them and try to grasp the reasons behind these laws! However, I'm not the type of person who would go after someone copying my texts! It would actually be the opposite, and a feeling of pride^^ It would just mean that I've done a good job^^ 😎

[hollow sapphire]

Tant que l'on se comprends, c'est suffisant^^ 👍

[charred merlin]

🥹 🇫🇷 ❤️

T'adore, mon ami.

[cunning mulch]

yeah... given, well, the types of songs I tend to 'write' (I use AI to generate a skeleton I flesh out), if you think it's banger enough, by all means share that sucker out

[hollow sapphire]

"Well, hold on! We only have one life! And in the end, it's so short that we have to put everything into perspective and not stress over things that don't really matter in the grand scheme of things."

[charred merlin]

spoke to someone about public domain
Meaning, asking whether or not you're allowed to upload public domain?

I explain it in the conversation around the post you linked, but while you're allowed creative use of public domain works, they aren't acceptable for uploads into Suno because you don't have exclusive rights to them (meaning you're the only person in the world who is legally allowed to use that clip)

Pretty sure in the Terms of Service, Suno encourages you to contact legal experts for up-to-date information regarding copyright law & AI because it's a fast changing scene, but as it stands, the Terms of Service that you agree to clearly states that you agree to only upload content that you're the sole owner/controller of (like songs you record yourself)

That being said, giving legal advice or clarification might not fall into their SoP, so unless it's a specifically phrased question, it may end up being low on the totempole, priority wise.

they should tell you upfront if you can't use other people's content
...they do, though? That's what the picture you replied to above states.

It's like if a restaurant puts up a sign that says "no horseplay", that covers "no running, no throwing food", etc

"You own (YOU made it) or exclusively control all rights (the person who made it isn't letting ANYONE else use it)" is PRETTY CLEAR and unambiguous.

That being said, if you don't agree with the terms of service that Suno asked you to agree to, you can use the clips on another service, or you can disregard the terms you agreed to and upload whatever you want "in spite of this prohibition", which very well may lead to your account being shut down, or potential legal actions. (the last part is less likely to happen if you're using public domain stuff, but this is SUPER early doors...)

[hollow sapphire]

**Only for your eyes: ** In France, works that have fallen into the public domain are no longer subject to copyright. This means that the economic rights (such as reproduction and representation rights) expire 70 years after the author's death, and the work can then be freely used by anyone without permission or payment of royalties.

However, there are two important points to consider:

1. Moral rights: In France, the author's moral rights, which include the right to the respect of the work and attribution, are perpetual, inalienable, and imprescriptible. This means that even if a work is in the public domain, it must be used in a way that respects these moral rights. For example, the work cannot be altered in a manner that would harm the author's honor or reputation.

2. New editions: If a public domain work is reissued with significant additions or a new presentation (such as a translation, adaptation, or critical edition), these modifications can be protected by new copyright. In this case, the rights apply only to the added elements, not to the original work itself.

In summary, while works in the public domain are no longer subject to economic copyright, it remains important to respect the author's moral rights and consider any potential new rights on recent editions or adaptations.

[cunning mulch]

So I can hit the ground running in two weeks, if I want to try and keep a song as-is while tightening up lyrics, is it better to give it the Instrumental 'stem' (which does have some ghost vocals in it), or the combined instrumentals/vocals?

@native rock and remember that a certain evil mouse company made their fortune off PD works

or the story of a based patriot that returned citizens' money to them from the government

[foggy wing]

My luck with audio upload so far has given zero results worth saving. I've found several extremely good samples in songs made by Suno, but when I trim them down and upload to Suno to get a new song with the earlier sample, it is just a bleak variation where the sample that was so good is now gone in the mix.

[leaden lark]

I believe it helps with the quality

[neat nexus]

Actually my understanding is that by changing the words enough they can end up with a new edition of the song which has its own copyright. Last year I made a Christmas album and I had to be really careful whenever I changed anything to make sure that I didn't also change that thing to something that had been done already to modernize it and was part of a new edition of the song.

It was rough...

This gets even more troublesome when the modern verisons of a song are the ones that everyone knows because they have supplanted the original in everyone's mind.

[charred merlin]

@neat nexus I thought I copied it, and it didn't copy. Please DM me the contents so I can paste in my own thread

Woodhouse is usually unhelpful in our interactions, but that was genuinely frustrating to have a structured guide automoderated because of "too mny emojis"

[neat nexus]

DM sent.

[cunning mulch]

hmmm.... well, what about using some other singer synthesis to generate my updated lyrics?

[vestal rapids]

you need training data sets, there aren't any

[charred merlin]

https://stability.ai/stable-audio ...from the same team that brought you Stable Diffusion, for a matter of fact...

[vestal rapids]

you essentially need a data set which has ingested Spotify

[charred merlin]

I've heard of people using external voice models to create vocal melodies for suno-generated songs, but admittedly I haven't looked too far into speech-synthesis, so @untold briar, our resident Bark Subject Matter Expert might be able to shed more light on competing services.

[cunning mulch]

I mean, in the end, suno is a tool for music creation, and one of many

so, moving between tools might be necessary sometimes to achieve what you're looking for

I mean, the song itself is already good enough I'm listening to it a couple dozen times a night

I've heard of some where you can change vocalists, but can it edit them?

[charred merlin]

Couldn't have said it better myself -- it's super convenient when a single platform can give you everything you want, but sometimes the burger is better from 1 restaurant, but their fries suck. The other place has AMAZING fries, so if you have "the perfect munchies" in mind, it might be worth hitting up both places instead of just eating a good burger and complaining about the fries 😂 ❤️

[golden rover]

Is there a way to remove some lyrics, thr song ia great but my vocalist have sang a [meta tag] 😅

[cunning mulch]

not yet

[charred merlin]

https://www.landr.com/fairtradeai/ is an attempt at compensating LANDR users for opting-into a program that allows their content to be used in AI datasets -- not free or opensourced like LAION for image-sets, but there are more and more training sets available 🙂

[golden rover]

How do you want to change the vocalist ? Cuz ive been changing mine often in tone, expression exemple

Seems to work well but it often stay in thr vocalist range of what the ai gave you vocalist

[cunning mulch]

same way you are: I've tightened up lyrics, and want to 'inpaint' the new lines

[charred merlin]

2 ways I can think of to handle it;

1 - create an extension point before the accidental lyrics, and continue from that point using only the correct lyrics (keep in mind that metatags over 2-3 words are more likely to be interpreted as lyrics in hallucinations)
2 - extract the vocal stem using the paid-Suno "Get Stem" feature, or use an external stem splitter and DAW like https://www.bandlab.com/splitter and https://www.bandlab.com/studio to remove just that section of vocals. (this way you need to take the song off of Suno.com, but it can still be shared elsewhere 💪 )

[cunning mulch]

be careful using the "Instrumental" stem from Suno: you will still be able to hear phantoms of the lyrics

[golden rover]

thanks 🙂

[charred merlin]

Look up the recent entry on "Terminator Extensions" in my pool party > #1256301881966198805 message

For your implementation, I'd want to place the terminator right after the vocals end, but leave as much context as you can before placing the pivot so that Suno has enough musical context in order to reasonably continue the musical theme and motifs, but you're limiting the "musical memory" to JUST before it knows what the vocalist sounded like, so at least Suno won't pull so hard toward continuing that specific vocal tone, and you'll have an easier time nudging it in the direction you want.

Not 100% guaranteed, because sometimes Suno just absolutely believes that it's making the right musical choice to "make something that sounds good, from a pattern-recognition standpoint"

I'm still working on making my rambles less ramble-y, so please let me know if there's anything you need clarification on 🙏 💙

[golden rover]

Nice, i did read is fast, i think i'll have to sit and try it soon, it could give interesting things

[golden rover]

thank for sharing with us !

[charred merlin]

Key takeaway is that "extend" can only use the clip that it's extending from to know how to continue -- knowing that, you can give Suno as much or as little "musical memory" as you want by using multiple extensions to choose the length of the "musical memory" you give it -- choosing WHERE you start and stop the context of that memory makes ALL of the difference 💪

[charred merlin]

You're very welcome -- have fun Sunovating 😄

[golden rover]

To be honest im pretty new here, and i dont understand all the Extend features. Im kinda still testing some stuff i found by listening other songs and on this discord

[charred merlin]

oh man -- it can be overwhelming getting started, some of my guides touch on intermediate-to-advanced level stuff, so think of the pool party more of "testing methods to grow into" instead of any form of "recommended reading" to get started 😅

#1151542418991239168 is a great place to dip your toes if you want to know how some of the features work, but in a few words, Extend lets you continue an existing clip, whether you just tacking an extra part onto the end, or starting from just before the vocals came out weird in that one verse in that one song that otherwise sounded amazing.

Worse comes to worst, feel free to ask any of the community-guides or the moderators, and they usually have a pretty solid handle on what to do (for all else, there's [e-mail] 🙂 )

[golden rover]

i did read alot, and im still testing what i did read haha, but i got some great result so far

[charred merlin]

I think I saw one of the guide-extraordinaires, @hollow hare, hanging around today 🙂

👋💙

[golden rover]

ill check it

[golden rover]

But the way i did change the vocals seems to work most of the time. But i'll have to test it more. Or i am just lucky, its sad that i forgot to remove a meta tag that isnt working well 'and it got sang, when i copy pasted my lyrics 😂 but damn so many info everywhere 😀😂

[charred merlin]

Ever since a ragtag childhood-friend group of adventurous scamps, not unlike The Goonies, managed to find and capture an actual Leprechaun the devs coded Suno, we're all able to access Luck™ though a subscription 😄

[golden rover]

haha

[charred merlin]

that was a weird typo, I gotta get this keyboard checked.

[tardy river]

I am trying to upload a 59 seconds mp3 audio, Suno complains "Failed to upload myaudio.MP3 Failed to fetch upload parameters"

[tardy river]

Ok, I renamed the file and it worked, it was a name like "sk30b_1835_19_35_forSUNO.MP3"

[charred merlin]

Thanks for following up with what worked for you! I've seen that issue present itself a number of times, and beyond grasping that it means that Suno didn't find what it was looking for, it fails to explain what specific parameters it failed to find.

Every little bit helps 😄

[golden rover]

ill post you the song, the tag seems to have work well this time, i know some have been ignore (1 or 2 tag in the song been replace numbers in the video lyrics) ill have to check, i must have use tag the AI cant process. the song isnt what i really wanted but im still in testing Phase (i'll have to Get a list of Comprehensive Tags for the AI to catch my ideas)
https://suno.com/song/e94a3a83-53db-4960-8b95-9f08748ce3a5

but i got some good result with some tag, and im happy it my learning so far 🙂

[charred merlin]

If this is the song where Suno sings the [metatag], can you share the timestamp of when it happens? 🙏 💙

[golden rover]

i think this one isnt

it was [energetic][add loud Electricbass rift] it was a guess , i saw that in another song haha

[charred merlin]

Ahh, I misunderstood -- this is the good result 💪

[charred merlin]

It's not so much a matter of "right" or "wrong", but I've noticed that metatags that are longer than 2-3 words are more prone to be sung as lyrics when Suno hallucinates, whether or not it leads to more hallucinations in general, I'm not sure.

Arguably, including any metatag is asking Suno to add something since we don't have the ability to "negative prompt" (https://discord.com/channels/1069381916492562582/1253341197917884530 though this IS a highly requested feature), so I wonder if [electricbass-rift, loud] would have been as prone to hallucinate?

[golden rover]

https://suno.com/song/b906fa6c-c1b7-4d23-ae39-fd92018fa2ac
i think my tag powerful is something Suno cant process
and [energic][add loud Electricbass rift], Tag is too long ive what i read

exemple for my wrong tags ( too long or did use uncomprehensive tags for the AI)

[charred merlin]

Even if it doesn't understand it, it's more apt to simply ignore it or use it as "prompt noise" instead of outright singing it -- unless the AI hallucinates / has been overloaded

[golden rover]

But when you check the lyric video, you can see what tag went wrong, they seems to replace it with numbers...a good pointer

but my problem atm is that im gathering so much info and alot might be upselet cuz alot dont work 🙂 but i have alot of fun

[heady sun]

In the "Style of music" box do you put commas between your words or not. I know the order kind of affects what happens right?

[golden rover]

Yes

https://www.suno.wiki/faq/style-and-lyrics/should-my-style-prompt-use-commas/

[golden rover]

Hope it will help 🙂

[charred merlin]

"dance, punk, guitar-solo" are 3 separate style prompts "dance punk guitar solo" is more likely to be interpreted as a single noisy prompt

Since it's hard to make "objectively bad" music with Suno, the result will probably still sound good regardless, but it'll be way less directed.

[heady sun]

Where should I stick a year (ex: 90s) in the style prompt?

[charred merlin]

lol, that's literally the next prompt experiment playlist, @cunning mulch was referencing a reddit post about it, and gave some methodologies to test, but haven't set aside time for suno experiments today -- if you're tackling it ahead of time, I can send the ideas they sent to me -- otherwise I'll give you a ping when it's out 💪

[charred merlin]

I haven't experimented much with it myself, but I recall some of the mods mentioning that prompt ordering doesn't seem to matter, especially if they're separated by commas

That was going to be one of the prompt experiment playlists, but if there's already a degree of community knowledge around it, I wanted to try to apply the efforts to something less documented before adding supplementary details to anything

I haven't tried the specific methods they shared, but I know that @naive quail is a mad scientist when it comes to playing around with lyrics, editing them, etc -- I figured that if there was anything "messy" like that, they would have stumbled across it and showed me something ridiculous he managed to science up 😛

[vestal rapids]

@charred merlin this just appeared in my Landr account, precisely what you were saying earlier

ironically I am pulling my music off Landr

[charred merlin]

you rock, dude 💙

In fairness though, there's a LOT that can be improved in my testing methods, especially as they're relatively simplistic and compartmentalized in the ways that I test things -- while it's easy enough for me to say "confirmation bias" without reproducibility following my "scientific blinders", there's SO MUCH we don't know about how Suno works, so we need people to do their own tests so we can try to address any blind-spots in my methods or points of view.

[cunning mulch]

because you don't want them used to train it?

@viral moss In my own notes, if you delete EVERYTHING in the 'Part 1', you'll cause it get utterly lost, and it willl absolutely start hallucinating

[charred merlin]

I believe it's opt-in, so they don't use your songs unless you want them to be included -- but this way you also earn 20% of the licensing rights 💪

[vestal rapids]

I have six albums on Landr, I am moving them because Landr hasn't been paying my PRO

[charred merlin]

I think it was just under a month ago they launched this, so it might be a rollout thing.

[vestal rapids]

This is the first time I have logged in to Landr this year

[cunning mulch]

@viral moss So, the ticket would be, if you're trying to edit some muffed/glitched/goofy lyrics, you probably want to delete only beyond from where you're extending from

I'm absolutely NO expert on that

[golden rover]

I found that if want punk metal work well but if you want a more blended punk-metal could blend the style more

[cunning mulch]

but if you held a gun to my head and made me take a guess, I'd say that the Extension uses what you've left in Displayed Lyrics as a guide to keep itself on track so you can be more flexible in choosing where to extend from

like, let's say you have a song, where you somehow put in "American as cherry pie," and thought that wasn't right

you go back, dutifully edit your Displayed Lyrics, and generate, but you clearly changed that to "apple pie," but all you ever get is cherry

[charred merlin]

"dance-punk"? Asking more as a clarifying question than bluntly giving an answer

[cunning mulch]

I think he's asking about something like "dance infused punk"

guessing you want punk sounds with dance rhythm?

or other way around?

or some kind of blend

[charred merlin]

Oh man, do you know the cult-comedy series Stella?

Only asking because you're niche and awesome like that.

[cunning mulch]

well, anyway... if you keep getting cherry when you want to change to apple, if you delete the line with cherry (and everything past it), and then generate your Extension, you'll find that you'll immediately get that nice warm slice of apple pie... and it might even throw in a scoop of ice cream on top

at least, I had that happen to me when I tried to modify lyrics in an Extension

[charred merlin]

Very first scene of the first episode, they're arguing about what music to listen to.

1 suggests "funk-rock", and the other suggests "funk rock".

An argument grows to "funk-rock" vs "funk-rock".

Sort of a tangential-aside, but it came to mind, and now I'm considering testing out dance-punk and punk-dance to see if the ordering within the specific mash-up prompts matters, as I've been told that ordering of 2 separate prompts doesn't matter, but this is something else altogether.

[cunning mulch]

that would be a hell of a thing to try and get nailed down

of course, you also have to deal with 3.5

[golden rover]

i would type it simple and Try it Prompt: 90's Punk song with dance Rhytym or Punk song, Dance beat,

[cunning mulch]

honestly, the biggest improvements v4 could make would be giving you better tools to look under the hood of your generations and see what it did, better tools to clean up songs you like but want to make truly shine without letting it go crazy with what you have already, and a bit more flexibility with structure

[golden rover]

indeed

[cunning mulch]

3.5 is great at coming up with 'full bangers' if you get your lyrics done well, but if it's not standard... oof

no joke, with my first 10 credits, I put some lyrics into the field, picked a couple styles, hit Generate, and got two bangers, that I had to just put to a vote

[foggy wing]

I commatize everything according to my plan. If I write "melodic death metal heavy drums" I get a mish mash sounding like the late 80's "Death" band. If I write "melodic death metal, heavy drums" I get In Flames, Arch Enemy etc.

[golden rover]

cause ai interpret as different variable and not just a mashup of 1 variable

[cunning mulch]

I'm still thinking about the BPM issue... what kind of tools exist for quickly analyzing the tempo of a particular song?

[golden rover]

i would like to know that too 🙂

[foggy wing]

Quick google, didn't try it: https://tunebat.com/Analyzer

[cunning mulch]

hmmm... key, BPM... wonder if it does time signature too

so, tools likely exist

[charred merlin]

That's usually what I use, as well as a metronome app

[foggy wing]

Ok, now I tried it. My prompt said "g-minor, mixolydian, 64/64, 60bpm", the tool said "D-major, 10B, 100bpm".

[cunning mulch]

What if, rather than trying to just directly set tempo, we instead compiled a list of BPM (ranges) for different styles?

[covert marsh]

Have you tried to ask ChatGPT ?

[cunning mulch]

...um what. Did I read that right? a 64/64 time signature?

[foggy wing]

Yes, it fills in the drum gaps and guitars beautifully

[charred merlin]

ChatGPT will confidently tell you EXACTLY how to make the song you want.

And Suno will respect almost none of it.

[hollow hare]

Not to mention it can lie to you with confidence

[charred merlin]

It's a great tool for making things easier, but it shouldn't supplant community knowledge.

[hollow hare]

ChatGPT: "St__r__awbe__rr__y has two r's"

[covert marsh]

ChatGPT isn't the be all end all but it's a good jumping off point. Always check for yourself.

[foggy wing]

This is what Gemini had to say:

I believe Suno is pretty well versed in the italian terms. If you use the english ones it takes over changing style of music.

[charred merlin]

I've had people create AMAZING songs and start to gain a minute understanding of how Suno works, then immediately fall into a pool of confirmation bias because "ChatGPT made a song better than I ever could", despite the fact that Suno didn't respect 1/10 of what it had been given.

"But that's okay, because luckily it ignored all of the stuff I didn't care about, and only chose to listen to the stuff that I actually care about." -- how do you try to engage collaboratively there?

[charred merlin]

These are WAY more effective than simply trying to command "bpm 90", but it's still good to note that it's all suggestions -- suno is a vibe interpreter -- if it's already going in a direction where "bpm 90" makes sense, it might not 100% give you that, but it can help things stay on track.

[cunning mulch]

right. I mean, I have a song that repeated lyrics, and even decided to end early... but I didn't care, because when it did do those things, well, it was utterly SICK in execution

[covert marsh]

I've been able to ask chatgpt to write the metadata for a song. I provide the lyrics. Most of the time I have trouble getting suno to be consistent with the output

[cunning mulch]

mine is telling me it's at 89 BPM

[hollow hare]

To add to this, the dance tags PDF linked at suno.wiki also had more Italian tempo listed. I took a screenshot and posted it at #📚┃suno-school message

[charred merlin]

That's likely because ChatGPT is misinterpreting how the metatags work in best-practice -- just being fed the information from the community wiki doesn't impart "tribal wisdom" onto it.

[covert marsh]

My problem

the problem isn't with chatgpt when I have suno create with the same instructions and it slowly shifts everything into a "pop" sounding song

[foggy wing]

When I ask for lyrical help, I usually write the lyrical theme in Suno, click generate to get the syllable format, copy it to Gemini and ask for an analisys to keep the syllable format while changing the words to my liking. I have been able to skip a lot of this by just asking Gemini to write the lyrics in haiku format to begin with. Suno LOVES haikus.

[charred merlin]

ChatGPT for Suno-specific "coding = 🤔
ChatGPT for writing lyrics = 👌

[cunning mulch]

https://suno.com/song/3800b7aa-66f1-4b21-8a62-b2f428871aed (Timestamp 2:30, repeated lyrics) (Timestamp 3:00, false ending)

[hollow hare]

That's 3.5 for you. If you are trying to use a genre that is not being produced appropriately by 3.5, make it in 3.0 then extend into 3.5

3.0 is better at interpreting the less popular genres while 3.5 is much better with the song structure

[covert marsh]

I'll have to try that. Thank you

[foggy wing]

Ok, I just had to try that about strawberrys.

[cunning mulch]

I wound up cutting that rest in about half for the version I listen to

[hollow hare]

Also you can try to specify for specific genres and see if they help with the style you want. I highly recommend using https://rateyourmusic.com/genres

[cunning mulch]

I also didn't ask it for the extra "(My guiding light!)" when that came up

but hell, I rolled with it

[charred merlin]

Ad-libbing isn't exactly a hallucination, or it's a very minor one at that, as it's trying to add familiar patterns that it recognizes in other songs that sound "objectively good" -- even if it's not our desired outcome.

[cunning mulch]

right

[golden rover]

Im browsing everywhere and i cant find what i need. anyone have a file / list of working tags ? I would like change how my vocal are in a verse exemple. But some tags i used doesnt work. Anyone have a list of that?

[foggy wing]

Just adding "3/3, epic, metal" makes everything corny pirate metal. A real mood swinger.

[cunning mulch]

those ad-libs turned out absolutely *siiiick* though

[covert marsh]

The thing I'm having trouble with in Suno right now is I want the music type to be like REDNEX. It seems to be confused with combining a heavy techno beat underneath the country/western sound

[cunning mulch]

so I didn't mind

[charred merlin]

It's not so much a matter of tags that are "guaranteed" to work in comparison to "wrong" tags, it's often a matter of how you're using them.

That said, there's a list of tags floating around here, I'll pull it up -- #📚┃suno-school message

[cunning mulch]

see, I guess I'm too used to trying to talk to it like it's a machine that understands technical terms and shortened tags

[charred merlin]

What are you using for style prompts? What you use as input is the largest determining factor on what Suno outputs.

[charred merlin]

https://tenor.com/view/let-me-be-clear-bart-school-say-the-line-bart-say-the-line-gif-4033914080036998615

"you don't "code" suno, you talk to it"

[covert marsh]

I've been writing meta data into the lyrics. It seems to get me better results leaving the style box empty

[cunning mulch]

oh lord... you say that, all I can hear in my head is the Simian Acres track from Blast Corps... but hey, that absolutely rocked, too

[charred merlin]

#📚┃suno-school message

If you like how it sounds, then that's all that matters, but it's hard to give actionable advice at the same time that "chatgpt does it better", even if the stuff that ChatGPT gives you is ignored.

[cunning mulch]

what would be a good way to try to then say, 'no ad-lib' without using a negative?

because I bet those repeated lyrics are also ad-libs that are horridly unwanted

what about a tag like [Unorthodox Song Structure] to warn it ahead of time?

[covert marsh]

I didn't say chatgpt does it better, you pointed me to someone elses comment. I'll try the Suno 3.0 and see if that helps me

[cunning mulch]

or maybe [Flexible Song Structure]? something to communicate to it to not be so rigid

hmm... I wonder, am I possibly causing it to try and create multiple vocalists if I give multiple descriptors for a vocalist over comma-separated style descriptors?

[charred merlin]

https://www.suno.wiki/faq/style-and-lyrics/duets/ not impossible, but notoriously difficult

[cunning mulch]

gotcha. I figured I was describing different 'parts' of a single vocalist, but wanted to make sure

vocal quality is kinda... I hope we can regenerate vocals in v4

[steel crescent]

Can someone help identify the name of the sound in the background, (Before the male voice) please? I’m not sure what it’s called and would like to reproduce this kind of ambient sound in other songs. I’m not trying to self-promote; it's just that my music knowledge isn't very advanced. That’s why I’m seeking help from someone more knowledgeable in the field https://suno.com/song/138f354e-33f0-428b-9e9b-a998ee166ac2

[mystic salmon]

Hey guys and gals, how can stop my songs from looping, I have to manually skip them when I listen to a playlist, I use chrome on my iPhone 11, thank you!!

[cunning mulch]

now... another interesting question that may have been figured out: When extending, if you leave the style blank, what happens? Does it continue using what it had before, or does it pick all new?

[foggy wing]

Ok, so I did a test with writing key, scale and tempo in italian.
So instead of "d-minor, myxolydian, 135bpm", I wrote "re minore, scala misolydia, allegro".
I did this for four songs and fed them into the key/bpm analyzer.
I then did the same again for two songs where I only changed "re minore" to "mi maggiore" (e major).

So I believe part of Sunos style is to knowingly always do a forced combination of minor/major and faster/slower, but only changing one of those at a part. So when the button is pressed it will make "minor, slow" and the other "minor, fast" OR it would make the first "major, slow" and the other "minor, slow". Or something like that, with reservation for prompt weights. What do you think @charred merlin ?

[cunning mulch]

what's 'Alt Key'?

[limpid badge]

Yes, that's the closest way to keep next part simillar

[cunning mulch]

okay, good. Picked up a new tip there

'cause I think I got a style shift in there when I kept the same style prompts

[foggy wing]

7A means D-minor has notes from A-major. A-major is then an alternate key for variation within the same key. I am not an expert on musical theory however.

[cunning mulch]

ah, okay

gah

[foggy wing]

This brings back high school memories... Oh, my god...

[cunning mulch]

See, the one I had to much trouble getting to generate was coming in as C-sharp minor (annoying bot), 12A in that tool

so, I suppose D-flat

so maybe this means that instead of trying to just control BPM, you can guide it into the two 'gears'

so, perhaps my song got guided into slow, major on that Extension

would help if I provided the song, huh?

https://suno.com/song/d844660d-0bf8-4951-b19c-fad7f6845a62

I had a devil of a time trying to get that darned chorus to actualy start, rather than have it repeat lines 7 and 8 of the first verse

eventually, I got that choked-off last word... bleh

that's why I'm so keen on somehow redoing the lyrics while keeping the instrumental

well, regenerating, anyway. I also have some actual lyrical touch-ups to get in too. Punchier lines, fixed-up meter, etc.

[charred merlin]

Turn your phone into landscape mode to disable the repeat button to the left of the pause button ❤️

[charred merlin]

It uses the existing song as style context and doesn't attempt to nudge it in a different direction -- it's not clear to me whether repeating the same prompts furthers the "style request" or if it's simply "ignored" because that's the existing style

[cunning mulch]

mmk... I ask because sometimes I'd get wildly different art styles with my Extensions, and they each went into their own music style

[charred merlin]

I was going to be a smart-ass and say "the one beside the spacebar", but I decided against it.

...I guess that counts as still making the joke, though.

[cunning mulch]

sometimes it would generate a denuded tree image, other times, it might make hearts in various styles, sometimes, it would make a very detailed map of Europe filled in with various textures

but each one veered off in its own way

did you see the South America heart? that one was funny

but you look at the pirate one's image (I think I may have picked that one out of the few tracks I generated because it just looked coolest), and there's no mistaking what it's showing you

OH... speaking of generated art... did you see what it did with one dude's track?

[charred merlin]

I think there's definitely an idea worth exploring, as I haven't spent much time considering the structures that Suno might be trying to apply -- I was just more aware that there WERE patterns it recognized, but hadn't considered them "knowable" to a degree.

Let me know if you have any ideas on how to go about testing it, and I'd be happy to collaborate & front the credits for testing 😄

[cunning mulch]

https://www.reddit.com/r/SunoAI/comments/1edm8j9/has_anyone_gotten_generated_art_like_this_on_suno/

I really wish I could see it in context on Suno...

lemme think up a test plan for that. what might be the best tag to use... [Shift Faster] and [Shift Major], maybe?

[runic ocean]

hey uh

has anyone had their music deleted by distrokid etc because it's AI music even if the lyrics are by you not by the AI (not happened to me yet I've just been warned since I'm in the process now)

like without warning just gone

no refund

just worrying if that's gonna happen

[charred merlin]

If you breach Terms of Service, they're within rights to cancel your service without providing a refund, but it's not so much as "deleting your music", as "not continuing to distribute it for you" -- you're still supposed to keep a master copy.

[runic ocean]

yeah

I don't think I've breached terms of service

[gusty acorn]

Why are you worried of breaching?

[charred merlin]

DistroKid is accepting of AI music, just make sure that you own all of the clips used, and make sure to credit any other contributors (like other artists who you extend from)

[runic ocean]

it's not that I've just been warned

am I able to post images here

[charred merlin]

If it's relevant to the conversation, yes

[runic ocean]

okay

[charred merlin]

If it's just someone telling you "be careful", it's okay, no need -- hearsay is hearsay.

[runic ocean]

okay

it's just a snippet from fb I was shown of someone's music from distrokid getting taken down by dk and people were saying it's because it's made by AI

I'm probably just freaking out for nothing

[charred merlin]

Context is everything.

Without knowing the entire story of that song's production, people will speculate and want to seem like FB experts.

[runic ocean]

yeah

how long have you had your music on spotify for?

[charred merlin]

Even watching a video of someone clearly being a jerk and attacking the cameraman, while the cameraman is acting super calm and concerned -- how do you know the cameraman hadn't been doing terrible things and antagonizing the person, then suddenly turned their camera on and acted surprised?

I know that's a super weird, barely related example, but without knowing the full context of what happened before and after, it's easy to sit back and act like an expert, but whenever someone shares their side of the story , there's ALWAYS an implicit bias.

If that's something that really concerns you, follow up with the person who posted it -- ask them to share the details of the song, to share the suno clips so you can look at the prompts they used, whether they owned all of the extensions or took "a cool beat and made it their own" from another Suno artist, etc

[runic ocean]

yeah

[cunning siren]

If you distribute do you have to state it‘s AI generated?

[runic ocean]

I was planning on adding that to my spotify artist profile when it goes live

[gusty acorn]

no

[runic ocean]

since I do wanna be transparent lol

[charred merlin]

I know when DistroKid drops you, they drop like an ex-lover who got their life together and needs to move on -- "sorry, this isn't working out for either of us. Unfortunately, there isn't more to say on the matter, and we wish you luck in the future."

[cunning siren]

Where did I see that kind of checkbox … maybe TikTok

[charred merlin]

instagram / youtube

[gusty acorn]

All music today is in some way made with help of AI. Plugins, mastering, autotune you name it.

[cunning siren]

True but „they“ want to be distinguished artists 😉

[charred merlin]

Absolutely can't hurt to list Suno -- I think some people are worried that simply using AI will give them a reason to be kicked off -- some people genuinely, intentionally falsify information to make it seem like it's not AI generated because they don't want to get kicked off because it's AI, ignoring the fact that they've given more reason to be kicked off for lying.

[runic ocean]

yeah

[charred merlin]

To anyone wondering if any service is right for them -- Read the terms of service, and decide if you want to partner with that platform. If you don't trust their terms of service, they probably aren't a good partner for you.

[cunning siren]

Truly not the first negative thing I’ve heard about distrokid …

[charred merlin]

People also knock on Dollar Stores, but they serve a hell of a purpose when that's all you need.

🤷‍♂️

[cunning siren]

The one and only negative thing that keeps me away from distrokid is their subscription model although it is tempting to use because it’s a very foreseeable cost per year no matter how often you distribute

[foggy wing]

I think these five prompts would suffice with the following lyrics generated by Suno (to let it build its proper flow).
When your playlist is up, I can do variations on it, to see if we get the same results with english music theory wording and also by numerical music theory expressions.

pop, Mi maggiore, 4/4, adagio
pop, Mi maggiore, 4/4, allegro
pop, re minore, 4/4, adagio
pop, re maggiore, 4/4, adagio
pop, mi maggiore, 4/4, presto

[Verse]
Music is just sound
Notes that dance around
High notes make you smile
Low notes stay a while

[Verse 2]
Clap your hands so loud
Hear the happy crowd
Drums go boom boom pow
Feel the rhythm now

[Chorus]
Twinkle twinkle high
Bass notes rumble by
Melodies that fly
Chords that never lie

[Verse 3]
Piano keys are fun
White and black and done
Strum the guitar strings
Let your spirit sing

[Chorus]
Twinkle twinkle high
Bass notes rumble by
Melodies that fly
Chords that never lie

[Bridge]
Every sound’s a friend
Songs that never end
Music’s in your heart
Let’s all play a part

[golden rover]

damn i got Nice result for different Shift in the Music and the Vocal but i cant achieve 1 thing, i would like my vocalist to Scream the Verses. Screaming like screamo (exemple SIlverstein - Bleeds No More) i have try alot of things, nothing that i know works
Prompt:
Epic metal, Japanese female vocals, dynamic shifts, powerful drums, intense riffs, Jrock,(aggressive Screamo Verse)
song:
https://suno.com/song/e1169267-abf9-4377-a4c7-72af2281f272
Anyone have an Idea ?? please ❤️

[cunning siren]

Maybe add „screaming“ to your verse-tags? [Verse, screaming female vocal]

[golden rover]

[intro] [Reverb vocal][Drum Beat]
痛みの旋律が
心に響くこの声
夜の闇に光を
希望を歌う

[Powerful Instrumental Transition]

[Verse 1] [aggressive Screamo vocal]
孤独な夜に 泣き叫ぶ
心の中で 壊れた夢
一人では 抱えきれぬ
痛みが続く 助けを求める

[Pre-chorus 1] [building tension][Dynamic Soaring Vocal]
心の断片を
(心の断片を)
集めて!
(集めて)

[Scream] A MELODY OF PAIN!

[Chorus 1] [Powerful, Expressive, slow Vocal]
孤独に苦しむ
この心で
誰かが必要で
傷を癒して
失った心
元に戻したい
心の断片を
集めて

[Instrumental]

[Verse 2] [aggressive Screamo vocal]
孤独な夜に 泣き叫ぶ
心の中で 壊れた夢
一人では 抱えきれぬ
痛みが続く 助けを求める

[Pre-chorus 2] [building tension][Dynamic Soaring Vocal]
心の断片を
(心の断片を)
集めて!
(集めて)

[Scream]
A MELODY OF PAIN!

[Chorus 2] [Powerful, Soaring, slow Vocal]
孤独に苦しむ
この心で
誰かが必要で
傷を癒して
失った心
元に戻したい
心の断片を
集めて

[Instrumental]

[Bridge] [Rhythmic Change][Dub Techno][dynamic vocal]
この孤独を
超えたい
手を差し伸べて
助けて
未来を取り戻し
希望を
心の破片を
つなぎ合わせて

[Pre-chorus 3] [building tension][Dynamic Soaring Vocal]
心の断片を
(心の断片を)
'集-め-て'

[Scream][Climax]
A MELODY OF PAIN!

[Bass Drop]

[Chorus 3] [Powerful, Soaring, slow Vocal]
孤独に苦しむ(しむ)
この心で
誰かが必要で(要で)
傷を癒して
失った心(
元に戻したい
心の断片を
集めて!

[Fade to End] [Powerful, Soaring, slow Vocal]
A melody of pain(pain) (x3)

here,s my Lyrocs and strcture

[golden rover]

i'll try that way to type it

nop didnt worked 😦

[cunning siren]

Is it screaming or was it a male voice? 😂

[golden rover]

the AI ignored it completly in 4 songs

[cunning siren]

That’s a bummer

[potent rock]

hey i have this song i really like, but i need to change one word. What would be the best way to do that?

[glad granite]

this might be too technical, but does anyone still uses the "Bark" text to speech model

[golden rover]

#📚┃suno-school message

[golden rover]

someone posted about that

[charred merlin]

The moderator Jonathan Fly does -- I believe he's said that he's accepting of DMs from users who want help, but just to be patient as it's something he's only able to help with in his free time. (also, he hasn't yelled at me for continuing to say this to people, so I'm gonna assume that it's cool I send people his way 😅 😛 )

I think you used to be able to use discord commands to get Bark generations the same as Suno/Chirp, but it's moved to the website now. So though this technically is still the discord where bark grew up, it's more of a passing relationship than an active subsection, but still neat to see people continuing to use it since it went open-source 🙂

[glad granite]

oh yeah

i remember talking to him once here itself

Bark is really good and powerful, however I want to know if there is any model that work in real time

[charred merlin]

https://elevenlabs.io/ ?

Along with Suno, ElevenLabs was named one of the 10 AI startups to watch in 2024 by Bloomberg

[glad granite]

interesting I wil try out the API

however the drawback in this is with the free tier its only about 10 mins of audio per month lmao

[glad granite]

https://tenor.com/view/spongebob-patrick-patrick-star-broke-poor-gif-14729256

[charred merlin]

In most cases in life, you can only pick 2 -- if you ever get all 3, hug it like the unicorn it is.

You've already picked "low cost" -- do you care more about quality or speed?

[glad granite]

idc about done quickly, just need low cost and high quality

[charred merlin]

The project management triangle says that's going to be low priority -- any company capable of doing it will have little reason to give their services for free, or they'll only provide the service on a trial basis to solicit customers

[glad granite]

i did found something called RealtimeTTS a free python module, but it requires high processing power which is not available on the device I use it on

[charred merlin]

Yeah, if you have the cost already taken care of as overhead in equipment that you own, like an RTX graphics card lets you make AI stuff on your own, so technically that's the bright orange triangle, "fast & good = high cost"

[glad granite]

yessir

might gotta take the things in my own hands lmao

[hot trout]

Is anyone else having issues with downloading songs from their Suno library, only to discover that the base volume is REALLY low, low enough to make the downloaded track basically unusable without using outside software to boost the volume? And if so, does anyone know how to fix it?

[unique stone]

If this is with the MP3, the file you download is the same as the file you play via the browser.
Last time someone had this happen, I had them send me the song link and the MP3 they'd downloaded and it was identical for me.
Possibly your browser is doing something to make the volume louder?

[hot trout]

Ignore this question on the grounds that I AM A MORON. Somehow accidentally tanked the volume in my player. Since I never change it on PURPOSE, it never occured to me that could be the problem. facepalm

I feel foolish. Thanks for the advice, though! 😄

[fallow wave]

was going to say, I get full bass freq in my songs.

[north sleet]

@charred merlin I build up from parts whenever I need to control Suno from being over zealous on the lyrics, BPM, or to keep the track clean and have less AI free will. I have also noticed the closer the custom lyrics/prompt gets to the 3000 character limit of v3.5 the more erratic it becomes.

By breaking the track into parts it keeps everything consistent. Example, “The Danse Macabre of the Undead,” was made up from three parts after I found the lead vocals I envisioned. Originally the AI was singing the lyrics super fast even though the BPM of the instrumentals was a slow waltz. Same thing happened with “Symphony of Shadows” in order for me to get the entire song closer to the target BPM I had to split into parts.

[delicate nebula]

How can I increase the sound quality and keep the background music the same, I really like the sound of this but it annoys me that the sound quality is bad? https://suno.com/song/700effcf-316a-4cf1-8a5e-29417cafd4e4

[charred merlin]

I keep getting distracted with new, cool things instead of making the original workflow guide for "breaking the song down into smaller chunks to exert granular control over Suno" that #1256301881966198805 was originally started to address, but you just hit the nail right on the friggen head.

Of course Suno is capable of amazing "one-shots", but if you have even the slightest bit of particularity for how the output actually sounds from beginning to end (beyond simply wanting it "to sound good"), it's practically a necessity to break things into chunks with extensions.

It breaks my heart to think of how many AMAZING introductions to songs have been thrown in the trash because they had "a bruise" early into them -- if you look at the original Suno clip for one of my favourite songs from my first album, on the website it's named "first 3 seconds" because I fell in LOVE with the first couple seconds before it went off in a different direction, and I wanted to build off of the cool part.

Just in the past few days I've adjusted some of my "scripts" to move away from suggesting that certain behaviours "make Suno hallucinate", but rather that Suno is going to hallucinate sometimes either way, because that's the nature of AI at the moment (or possibly because Woodhouse spilled some liquor on the circuitry, it's hard to tell for sure), but instead of being a black or white situation, there's a lot of factors that can contribute to hallucinatory behaviour, eg. approaching the character limit, stuffing too many words inside of a metatag.

You made my night posting this 😄 ❤️

[charred merlin]

I'm gonna try something -- gimme a sec

[charred merlin]

https://suno.com/song/8fd5973b-79fa-4a79-b5ce-010c636d0d90 👍 / 👎

In either case -- if the original track was 5/10, what do you rate the new track?

[delicate nebula]

Honestly, there is no difference. There is this background noise that really bothers me.

[charred merlin]

...allow me another second to confirm something.

[delicate nebula]

Thank you for your effort. And the song is really great, I don't know if you like it.

[charred merlin]

Little do you know I'm walking around my apartment singing "menjam seeeeee" without knowing what it means.

[delicate nebula]

haha
That means,"menjam seeeeee" I'm changing

[north sleet]

Hallucinations can be happy accidents too.

[charred merlin]

It might not have resolved the issue that you were hearing, but I was relatively certain I was able to hear a difference -- borrowing a technique that I learned from the guide, furrybeans, I placed your original track, and the version that I generated side-by-side in Audacity and inspected the difference between them.

If the songs were completely identical, the "difference" would be absolute silence, but here is the resulting track.

As you can hear, the "difference" track is quite noisy, meaning that a lot of detail had been added in the second version.

[delicate nebula]

Yeah i can hear that, great analysis!

[mystic salmon]

How am I not seeing this? Lol thanks for your help and patience 🙏

[north sleet]

Looks like they changed their website player again. The loop button used to be right before the three vertical dots.

[charred merlin]

Looking into this now

[mystic salmon]

Yeah I remember I had changed it to loop before version3.5 but I figured they had placed it somewhere else lol 🤷🏼‍♂️

[mystic salmon]

🙏

[north sleet]

Though now that I look at it the player the left track skip, play button, and right track skip used to be positioned on the bottom left.

[charred merlin]

What device / browser, etc?

[mystic salmon]

Is this on a computer?

[charred merlin]

iphone 12, edge browser -- not on desktop view

[mystic salmon]

IPhone 11 chrome

[mystic salmon]

Let me try edge

[charred merlin]

jeez -- edge is just a stylized chrome... something's going on here.

[north sleet]

iPhone Safari looks the same as the Chrome browser.

[charred merlin]

...and Safari is non-chromium based

https://tenor.com/view/tom-delonge-blink182-wtf-meme-what-the-heck-first-date-gif-25383351

Have all of the mobile browsers gone mad?!

[north sleet]

Even requesting the Desktop version of the site on the mobile Safari or Chrome doesn't solve the missing button.

[narrow cypress]

anyone else play with bilateral prompt at all? With much success? This is one of five that caught on to the concept:
https://suno.com/song/25ada1be-55f5-4fe7-b2a8-39eef16b0802

[charred merlin]

@north sleet @mystic salmon mobile bug confirmed -- you guys rock 💪

Google chrome on mobile doesn't show the additional playback buttons

[north sleet]

Even more mystery is that Safari on iPad shows the player without issue. So, it seems to be isolated to just the mobile device.

[mystic salmon]

With edge, my problem resolved! Thank you my sweet miracle workers!

[charred merlin]

I made a post in #🪲┃bugs-web, but we need to wait for the mods/guides to add it to the official #1252292428258480209 list -- thanks for your help researching this on multiple devices 💪 Many hands make light work 🙂

[mystic salmon]

Any idea when the suno app will be available in Canada? Lol

[mystic salmon]

❤️‍🔥

[charred merlin]

https://tenor.com/view/soon-soon-meme-batman-soon-soon-batman-batman-gif-26122636

[mystic salmon]

🤙

[charred merlin]

(tbf, it's nice for listening to music, or letting new-to-casual users create songs with little-to-no friction between having an idea and hearing the song, but intermediate-to-casual users will get more out of the web-app, as the iOS app is pretty feature-light at the moment; no extend, no reuse prompt, no model selection)

[charred merlin]

iPadOS is basically fancy iOS, but it's a LOT closer to MacOS in terms of Safari -- but STILL a mystery

[north sleet]

I use the browser to create either on computer or iPad. The mobile app and browser seems to cramped to do anything on.

[mystic salmon]

Ok good to know!! Thanks again! 🫡

[charred merlin]

Many welcomes from the Atlantic side 🇨🇦⚓💙

[mystic salmon]

😎🤘

[north sleet]

lol iPad OS is nowhere near macOS. It is just a glorified version of iOS as you said.

[charred merlin]

I stopped working for AppleCare around the time that MacOS Monterey came out, so it's been a while since I've been "forced" to keep up to date with all of the extra features and whatnot, I'm not quite sure what advancements MacOS-Safari has made, but I remember there being quite a lot of different ways to use Safari on iPadOS -- either way, until iPadOS manages to properly scale all iOS apps, I'm always going to throw a bit of shade at it.

[north sleet]

I am using StageManager at moment and it sometimes is nice and other times is frustrating. I wish they would add arrow keys to the on screen keyboard. Touch/Pencil to get precise doesn't always work as expected.

They need to make the iPad Pro the hybrid version that runs macOS with touch capability.

[charred merlin]

oh geez now you're talking

I'm sure we're not that far off, given how I expect they're going to leverage Apple Intelligence. I'm so salty that got pushed back from iOS 16 launch (convincing myself that it was ever part of launch plans)

[north sleet]

Apple says they have no plans to put macOS on the iPad according to an interview with Craig Federighi during WWDC.

I am not too impressed with Apple Intelligence yet.

Conversationally it is just like talking with CoPilot on Bing. Apple should have partnered with Inflection AI instead of OpenAI as Pi is so much better conversationist.

Though I am bias since Siri doesn't know where to hide the bodies anymore or even open the bay doors. Apple has stripped all the good easter eggs out of Siri.

[charred merlin]

It's been a while since I've checked back in with Pi, but it's hard to deny just how impressive the level of emotional intelligence Pi was able to convey in comparison to the competing frontrunners in "the early days" (oh my lord, I'm talking about the last 2 years like I've aged 20 within that span)

[charred merlin]

If you're a Harry Potter fan, say "Lumos" to Siri.

"Nox" for the opposite

"Accio {app name}"

[charred merlin]

Sort of a related tie-in, the dark conjurers who summoned Suno from the eldritch depths founders used to work for an AI company called Kensho, and something that company is actively tackling is how to mitigate toxicity in LLM-interations https://blog.kensho.com/an-introduction-to-mitigating-toxicity-in-llms-pt-2-fc67d70c8acb-- the Siri body-hiding thing is actually a great example -- while the developers thought it was funny to have Siri direct people to actual drainage ditches, someone actually followed Siri's advice and it was a whole thing.

That was back when it was a specifically programmed choice by the developers to have it interact that way. Now developers need to worry about people tricking Siri into giving specific details about what to do if there aren't drainage ditches available and you only have 15 minutes before the police show up. Beyond just chatbot-based interactions like preventing idiots from learning how to make napalm, this is the sort of stuff that we need to explore before we can start to introduce AI to capabilities that need a sensitive hand, because unless we have people researching the edge cases of these "hallucinations" and how people can take advantage of them, bad-actors will always have a one-up because they actively seek to break systems and gain any advantage they can.

[north sleet]

Just have to remember that when I am in a bind and need the magic want to show some light.

[north sleet]

There is always going to be someone that ruins the fun for everyone else.

[charred merlin]

I know it can be subjective whether it sounds better to one person or another, as it may not have removed sounds you didn't like (just simply added more sound on top of it, I cannot be certain), but an explanation of what I did so you could try different versions some time to find version you prefer 💪 🙏

I saw that your song was generated with v3, so I extended onto the end of the song with v3.5 and simply used the style prompt "end" -- one of the generations I made was 0m 39s, so I clicked "Get Whole Song".

Furrybeans had previously pointed out to me that Suno seems to "upscale" songs when you extend the song using v3.5, so I wanted to test if extending the end and getting the full song as a v3.5 track would increase the sound quality of the earlier sections -- I believe it did! 😄

I will send you the .mp4 and .wav of "Menjam Se" so you can use those tracks as your own, if you should care to ❤️

[delicate nebula]

Wow man, you really tried, thank you very much for your time and patience to deal with this issue. Maybe the V4 will solve all the problems with generally low quality sound. 🙂 💯

[charred merlin]

comparing the difference in audio quality from v2 > v3 > v3.5, I can only imagine!

[honest vessel]

how do i delete my account

there is no option

someone help

[charred merlin]

You need to email [e-mail] 🙏

[honest vessel]

thanks

how long will it take

[charred merlin]

The support advisors would be in a better position to answer that, I'd include that in your email to them. 💪

[honest vessel]

thanks a lot

[errant matrix]

Hey, does anyone know how to make a contest?

[amber vigil]

anyone in here know how to successfully pull off mumble rap? i've tried a few styles but suno doesn't understand / execute them. I wondering if anyone has seen it successfully done on suno.

I am thinking i might just go in phonetically to replicate it

[north sleet]

I am not familar with that genre. Do you have any example artists?

[amber vigil]

uhhhh can i share a link in here?

a youtube link?

actually i don't feel comfy sharing that link

if you search mumble rap examples - you can hear them - basically they mumble words

usually they have strong bass beats with lower lyrics and great rap flow

[north sleet]

If you don't feel comfortable linking the full YouTube link you can just post the characters after watch?v= in the URL. Then we can just paste that into YouTube to see the actual video link.

In order for the AI to mumble words you would have to spell out the words so that it is not able to articulate the pronunciation, which to us would make it sound like it is mumbling.

How to do that is the tricky part.

[amber vigil]

yeah that's what i'm thinking too

the ai doesn't seem to know what mumble rap means

atp

[random solar]

Hi, I'm new here, I wanted to ask if its possible to download only instrumental in a song I created (I only find button to download full audio)
I tried using external audio splitting applications, but it still have a hss sound due to suno's special vocal

[wind breach]

Hello Dears all, I have question regarding generation of RTL lyrics on Suno. They are all scrambeled together and it doesnt look like suno is able to recognize the RTL text, Like arabic. What to do in this situation ?

[untold briar]

I think RTL languages can work in Suno. Suno can scramble lyrics in any language. Usually this is called 'hallucination'.

If RTL languages did not work at all, I think I would have heard more user complaints. There is an Arabic channel here https://discord.com/channels/1069381916492562582/1155315011519004733 but it is not used very much. Maybe you can check for working song prompts in that channel. Or in other RTL languages, there is also a Persian channel for example.

[wind breach]

Thank you very much for your response. Im very happy with suno, and I can create songs in many RTL languages, even ones that are rarely used online and rare dialects. Its just the lyrics during the video generation

[untold briar]

Ahh, for videos, that may be a Suno bug, yes. I thought you were talking about the singing.

[wind breach]

yup, thats the main issue, because when you showcase a song, its nicer for it to have synchronized lyrics

[untold briar]

In case you missed this feature, on a song page you can click "Edit Displayed Lyrics". Then you click "..." and "Regenerate Video" to recreate the video with the new lyrics. You may be able to test different text formatting and find something that works.

[wind breach]

thats very interesting! yes I did not know, thanks ! I will see if that helps

[untold briar]

I think it's a "Pro" only feature (any plan that isn't free) though

[wind breach]

I did purchase suno, but I dont see it unfrtunately

[untold briar]

It only shows up on your own songs, on the single song page. Like this

[wind breach]

Yes, thank you. I found it but still that doesn't work. I think at the moment Suno is unable to include RTL lyrics in a Video

[tulip harness]

how do you stop v3.5 from repeating verses from before the extended part?

[north sleet]

You either get lucky or you have to build the track verse by verse extending out.

[north sleet]

A test that turned into a full track. Sip'n on tha Miix is my attempt at mumble rap. https://suno.com/song/3e496d75-ece5-4757-a974-209d61e4bb98

[north sleet]

I elongated the words to try to get the AI to slur or mumble the words. For example the first four lines of the lyrics.

Yeeaaah ... I got the drrip ... Diamonds... on my riiiist, uh
Pawll'up ... in the whip... yeeaaah, sippin, sippin on tha mix, uh
Eyesss feelin' heavvvy pawkets full of green yeeaaah
Talkin' low and slow (uh) 

[shut moat]

so i have discovered that if you write your own lyrics it tends to lean towards speaking instead of singing, it's just me? writing too many words.

[shut moat]

why'd you want mumble rap?

[north sleet]

@amber vigil was asking if anyone successfully pull off mumble rap.

[shut moat]

I just tried removing all the spaces in my lyrics.

didn't work.

https://suno.com/song/71673b04-877c-4297-9b47-3e4449c60439

not this version, I already deleted the songs, it butchered so many of the words.

[unique stone]

Seems to depend on the length of each line of lyrics for me. If I want it to speak I usually type full sentences or even a paragraph without starting a new line.

If I want singing, then I try to keep the lines short. Break sentences up across several lines, at a point where it seems like it'd be natural in a song.

[shut moat]

for this one the spoken word works, I also used beat and slam poetry in the musical style choice.

[north sleet]

I originally started with the first four lines above and then expanded to more lyrics so I didn't hear the same thing on repeat each generation. Diamonds started out as Di'munz and then went to Diiyaaamonds, but it sounded weird. To much expansion the AI just butchered the words and didn't flow between lines.

[unique stone]

Some of the lines here are long for lyrics. An example of what I'd change is to format it like:

sprinkled with a dash
of a failed weak attempt
to restore faith in our species

Or

sprinkled with a dash of a failed weak attempt
to restore faith in our species

Breaking up the lines like this can help to indicate the flow of the lyrics so the AI can figure out what musical structure should accompany them. If it looks more like a full sentence of text it'll think the flow is more like that of a speech.

[north sleet]

I try to shoot for internal rhyming, which helps for longer lyrics.

[austere minnow]

In Suno, does the separation of voice and instrumental work well or does it leave artifacts?

[north sleet]

All stem seperation processes either it be by AI or software leave artifacts that require cleaning up. Some do a better job than others in helping in moving the artifacts to the correct layer like SpectralLayers (Stienberg $79+) or RX 11 (iZotope $99+).

[naive quail]

Ya, I think spoken word does work well for that. As @unique stone noted, it definitely seems to be related to the structure of the lyrics. If I change the style to Acoustic Folk, Upbeat, Dobro, Resonator Guitar, Happy, Joy, I get much the same result. But if I then throw in a bunch of line breaks, Suno makes more of an effort to sing it.

I dig the way this one came out https://suno.com/song/9741917f-ad0a-451a-82d8-c05352035d8e

[charred merlin]

As the devs admit in their accountment post, it's only v1 of the Stem-Separation feature that they "squeezed" out of the v3.5 model -- that disclaimer aside, there's a lot of room for improvement, especially if you've used other services for comparison (https://www.bandlab.com/splitter, for instance), but for people who aren't technically-savvy, being able to make a "karaoke version" of a song they made without having to take their song outside of Suno must be really exciting, regardless how perfect it is to start 🙂

[charred merlin]

dear lord, I didn't realize how much I needed resonator guitar to start my day.

[naive quail]

guitars are always the best way to start the day... Then Espresso. 😄

[narrow plaza]

Hi! Do you know if its possible generate evil voice ?

[charred merlin]

After some of the things I've heard, I'd be more curious about what you CAN'T generate through Suno... 🤔

[narrow plaza]

I have try many times but unsucces

[charred merlin]

Not meaning that there's always an easy or reliable way to get what you want explicitly through "obvious prompting" -- Suno is a curious new instrument, and we're still figuring out how to play it most effectively, but there certainly are people who try to get specifically evil voices. @naive quail and @unique stone put a lot of their experiments on public playlists in Suno 🙏

[charred merlin]

Are you simply trying different variations of "evil voice", or have you attempted to reapproach the issue from different angles?

[narrow plaza]

I have try
[Marc]
[Male voice]

And
[Death]
[Evil voice / dark voice / deep voice...]

[charred merlin]

If I had to guess, using Simple Mode to say "make a 1950s sock hop song" is vague enough that when it tries to "fill in the blanks", it ends up making something more or less similar to https://youtu.be/U3KjDpvmtwU, so it stops itself because it has been trained specifically to avoid copyright infringement.

This sounds more like misunderstanding why the song was being flagged, until it eventually generated an unlikely example that didn't blatantly infringe Danny and the Juniors.

[open sleet]

The style tags matter a lot as well, alongside the metatags

I have found recent luck of grabbing style tags straight from ai such as claude and that, might have some luck potentially

[vestal rapids]

are you a free user?

then I don't understand the problem, you get two clips to audition

[charred merlin]

Just including "voice" is likely enough to push it into a generic direction.

Instead of saying "evil", which is very vague and is interpreted a specific way by yourself, try using thematic descriptors for the voice you imagine , [deep], [growled], [baritone], [snarling]

Keep in mind that EVERYTHING you use, lyrics / metatags / style prompts, all influence the song, so you might need to be very careful and methodical in your use of extensions to control the musical context that Suno pulls from the songs you extend from (see the pool party for a guide on "terminator extensions" for an intermediate/advanced read > #1256301881966198805 )

[sweet parrot]

hi guys where is the key ?

[vestal rapids]

there is no key, this is a music discord server

[naive quail]

But what IS evil? 😄

[sweet parrot]

Why it brings me here then XD

[naive quail]

https://suno.com/song/135504dd-0dc6-4ec0-9d75-dbbdf0118622

[charred merlin]

Because it's wrong.

[sweet parrot]

k XD

[charred merlin]

https://discord.com/channels/1069381916492562582/1254418510885687366 It's a known-suggestion, unfortunately the post has expired so you can't upvote it, but it's on their radar.

There's actual reasons related to how generative AI works, but if you're simply focused on having the devs understand that you'd like variations, or don't always want 2 songs, the post I linked covers that -- or you can go to #💡┃suggestions-web to continue the conversation and let them know that's something that you're still really passionate about seeing in regards to changes you'd like to see made.

[narrow plaza]

Thx bro

[neat nexus]

https://tenor.com/view/rick-moranis-are-you-the-gatekeeper-ghostbusters-horse-gif-9322927

[neat nexus]

Because the guy who put together what you're looking for flubbed the link. As a result I'd be really really leery of running any code he's made...

[winter thorn]

https://tenor.com/view/ghostbusters-sigourney-weaver-are-you-the-keymaster-keymaster-are-you-gif-24175104

[charred merlin]

When I realized what "keymaster" and "gatekeeper" were euphemisms for, the movie becomes a lot dirtier in retrospect.

[charred merlin]

🙏 You're very welcome ❤️

I always try to think that the more specific of a voice or a song that I want, the more time, credits, and creative approaches I'll need to put into getting it.

That said, sometimes you "stumble across" really cool stuff when you're not even trying to, and getting a crazy terrifying voice by accident while you're experimenting with other random ideas makes for a GREAT intro track to extend from to "use the existing demon voice" 💪

[charred merlin]

Hell yes -- a funny greenleaf

Welcome to the server, @winter thorn!

Thanks for stopping by tech-chat, we know you have your choice of weirdos to hang out with, and your selection is objectively correct. ❤️

[signal sentinel]

Hi there
I'm trying to use the upload file option for the first time and every time it says upload failed. I uploaded different fragments, observing the rule from 6 to 60 seconds, tried different formats, mp3, flac, wav, and it still gives an error. What could be the reason?

Thank you, but unfortunately it doesn't help (

[uneven crane]

I have same issue right now

[signal sentinel]

Got it, thanks, maybe the general server is down. I'll try later then.

[charred merlin]

#🪲┃bugs-web message Might not be EXACTLY the same issue, but the devs are aware of issues regarding uploads, especially relating to the "failed up fetch upload parameters" error message -- no estimated turnaround time, but it's on their radar 🙏 ❤️

[charred merlin]

Can you let me know if you're still getting the same error message?

[stuck plinth]

Any update in the 403 error in the known bug section? I haven't been able to use Suno for over 2 months now 😦

https://discord.com/channels/1069381916492562582/1252332553738653838

[charred merlin]

The Discord server is more of a supplementary support channel for Suno while their newly hired support lead starts to feel things out -- it used to be that there wasn't really an official support presence on discord, but it's starting to enter fledgling stages.

For now though, the best and most direct way to receive official support is to email [e-mail] for general inquiries, or [e-mail] for account-related inquiries. 🙏

[winter thorn]

💖 thank you! the information here by the weirdos has been very enlightening

[leaden lark]

How’s tech chat going today?

[charred merlin]

MUCH more cognizant and lucid than 12 hours ago 😂

How was your shift?

[leaden lark]

Not too bad, still had issues after power went out at work (though it did come back)

[charred merlin]

"Why is it that the power goes out whenever Sara works, and why is there always vandalism at the electrical fusebox?"

"Who cares, she's making some AMAZING albums in the extended break time, and to be honest, jamming to them is the only thing keeping me coming back to work at this point."

[leaden lark]

I have ideas for songs but I need to wait a little longer for my refresh

[charred merlin]

https://tenor.com/view/scrubs-milos-anguish-screaming-rain-gif-23249196

"Why have you forsaken my creativity, credits?!"

[leaden lark]

True, I’m sitting her at 5 credits

[red elbow]

https://suno.com/song/80a11ad1-aec5-4a31-9b20-db061522a89c How can i get something alike this? I managed it by random once, and never again, i've tried every word combination i can think of

Suno has an incredibly infuriating habit of putting an COMPLETELY UNFITTING background melody to damn near every song it makes and it is driving me insane

Such a waste of potentially good tool

[charred merlin]

(blatant copy-paste from an ongoing DM chat regarding a similar matter 😅 )

I find that v3.5 tends to drift toward "generic" if not given strong, effective use of prompting (similar to how old image generations needed simple prompts, but newer image-generation software thrives with more descriptive prompting)

If you're looking for more broad & versatile outputs, v3 really excels, while v3.5's strengths lie more in the song-structure and musical cohesion side of things -- both incredibly powerful, but just excel in some areas and lack in others. 💪

[untold sentinel]

Hi!

I have a question about mixing different music genres.

e.g. if I create a song with "pop, rock" style. Is it the same than say "50% pop, 50% rock"?

How could I achieve to move the "needle", so I could create a song "20% pop, 80% rock"?

I suppose natural language could help here (please confirm I am not wrong :)) but what happens if I select my own lyrics?

Sorry if it's too tricky or even a stupid question 😄

[charred merlin]

No such thing as a stupid question if it's asked with pure intentions ❤️

[mild rain]

can anyone give me some pointers I am trying to create a nice emotional ballad to a song I wrote but no matter what prompts I use it all seems too pop or rock based and the vocals are always some low singing dude when I would prefer a higher key sang in more of an emotional soft way, Like a proper ballad.

I have spent 4000 credits today alone just trying to get something I like and in the style that is in my head. Any pointers would be so welcomed right now, I seem to just keep generating the same stuff with the prompts I am using, even though I am changing things around constantly.

[hollow hare]

Not a stupid question 🙂 There is currently no way to define how much of one genre is in a song (though you could use vastly different genres using some tricks like terminator extensions Harber speaks about at https://discord.com/channels/1069381916492562582/1256301881966198805)

What you can do is create fusions where you can define one genre as the rhythm and the other as the melody for example. Here is a playlist where I tested that with a cappella & dance pop: https://suno.com/playlist/d83502d8-c487-4c95-82cc-412bc2823288

As you can see, there were three primary tests: inclusion and exclusion of the fusion keyword, swapping of the rhythm/melody keywords, and different genres (I usually used a cappella because it was easy to tell when it is used)

You can also try to look up if there is a specific genre you want at https://rateyourmusic.com/genres. Such as Pop Rock (I do not know how that would differ if you used Rock Pop, hmm I should test that later)

If you were unaware, using commas in the style prompt does effect the weighting. Each set of keywords, delimited/separated by commas, are given equal weighting while the weight of each keyword in each set is heavy towards towards first of the word and decreases thereafter

[hollow hare]

Try starting it off in 3.0 which is better at understanding more diverse genres (I recommend making it as an intro instrumental). Take a look at rateyourmusic genre list (linked above) to see if there is a specific subgenre for you to use.

After you get the start you liked, extend part 2 as 3.5 which has a better song structure.

[charred merlin]

@hollow hare -- why did you never tell me about this testing playlist?! Can I post this in the pool party?

btw, I really vibe to this https://suno.com/song/9714d57f-9edd-4b48-9371-ab0f4412be84

[charred merlin]

I have no idea how to properly explain this, but it reminds me of "the scary part" of Home Alone after he hides in the nativity scene.

"This is my house, and I need to defend it."

[hollow hare]

If course, go ahead 🙂 I have three test playlists:

• The fusion one you seen

• Accidental duets which is not so accidental from like the Big Mac Dreamin' test series (still trying to figure out how to get more consistent vocal line swap and group duet): https://suno.com/playlist/11ff3827-bdcf-4ef5-bd04-da8647c6cdb6

• The most useless one was trying to shift the tempo (https://suno.com/playlist/2b851ed4-21f4-4c9b-bbdf-e93840509d7b), barely spent much time on that

[charred merlin]

https://tenor.com/view/home-alone-this-is-it-gif-15171445

re: the tempo playlist -- did you find that the testing methodologies could have been improved upon (removing unnecessary variables like thematic lyrics that could impact the style, like my "I'm so angry right now" lyric-only track?), or just that the theory was disproven?

I'd argue that it's just as important to share "negative" results as it is to share the positive ones

[hollow hare]

Hmm I didn't test that but I would not be surprised if that would work, as well as removal of a specific genre definition (I have a feeling that the genre used does constraint the BPM which is easy to tell with genres like breakcore which have very fast BPM)

Tbh, the only reason for that tempo test was because I was trying to get a fast rap style because someone was trying to achieve that style

(they were trying to achieve it to sound like Eminem hence the inclusion of the Canadian Vocals part)

[golden rover]

Hey all 🙂 i did alot of test and i cant get the voice delivery that im looking for..
Im looking for a singing delivery like in the song "akatsuki by babymetal"
Or to achieve a vocalist that is singing slower than the song fast rhythm in general.

Any prompt idea or tags to add to lyrics?
Cuz it often want to sing to an faster pace in verse, and actualy sing in the chorus..

Im open to suggestion of option to try, im starting getting a headache trying lol
😁 thanks all

[cunning siren]

I got good results when i stated to have slow melody with fairy voice

[golden rover]

Humkay

[cunning siren]

You may also achieve good results if you ask for lullaby like piece

„A dreamy lullaby sang by a fairy“ for example

[leaden lark]

I haven’t used lullaby before, but I have noticed chill vibes for some of my stuff

But using atmospheric in the style adds a very relaxed vibe

https://suno.com/song/b6a3b1d1-1899-49dc-9636-0f036b261b85 for example

I’ll test out lullaby when I can

[violet mesa]

This morning was one of these rare days, sunos songs came pretty good out. I've discharged only 4 songs , the voice was too noisy, no way to clean that up in my daw. It's a shame, cause the music was pretty awesome..
Well, from 20 + songs I choosed 8 , and have it published..
Every song was processed in my daw... only some EQ and hall and normalisation was needed.. came out pretty cool. #musicproduction #daw #homestudio #indiemusic

https://youtu.be/XRnECt-lA-w?si=Cw8pDT7nk38A_q4z

[amber vigil]

nice work!

[tight spade]

@charred merlin hey man you around?

@charred merlin can i send you a song

[charred merlin]

hey dude -- you're always free to DM songs! I might not always be able to listen to them right away, but I always try to get around to them ❤️

[tight spade]

sent when ever you have time

more looking at the structure of tags and such

eek

[wet minnow]

🧹 had just this post like 10 times. qualified for a ban. thanks for spotting it.

[golden rover]

Is it possible to make a song with only tags in lyrics and no prompt?

[charred merlin]

https://suno.com/song/65a2e520-e54a-4def-be67-d483a08a49eb

[charred merlin]

I'll get around to listening to most songs that are sent to be eventually, but I really enjoy when people let me know why they are sharing that song in particular -- you have made hundreds by now, why is this one special to you? 📊

• Do the lyrics have special meaning to you? 💛
• Did you write it for a special event in your life? 🎓
• Did you try very hard to achieve a specific sound, and are pleased with the results? ⚙️
• Is there a part that you find very powerful? Ex. "the guitar solo from 1:30 ~ 1:50" 🎸

This takes it from simply asking "if the song sounds good", and turns it into a musical conversation 💪 🙏

[tight spade]

for sure

[golden rover]

Ok ill rectify my question, if anyone made a song, with [TAGS] in the lyrics part that have really modify the rhythm or even the style of music of the song or any variation that have work with tags?
If yes, im open to receive your song in DM, or dm me to discuss about this 🙂 thx in advance

i would use em, i'll just study the structure of the changes

I dont want to spam this thread with this. Im just trying to learn more

[north sleet]

Thanks. I am sure you can make one that is better. It was a new genre for me.

I have a feeling my Spotify’s Discovery Weekly on Monday will have some mumble rap in it as I listened to at least seven different artists to understand the style.

[charred merlin]

SixStringToker has a bunch of public playlists, and he's a pretty prolific Sunovator; https://suno.com/playlist/8fc9cb0e-1a70-40e7-92b7-0f762dd2c109

[charred merlin]

Heck, spam away -- it's been a quiet night in here besides that crypto-spammer dude getting banned 😛

[north sleet]

Sounds like another idea for lyrics.

[amber vigil]

i think you did really well tbh!! I doubt I will go to such efforts but maybe i can talk my friend into doing one with me as a fun bonding exercise.

Thanks for showing how it can be done. ❤️

[golden rover]

@charred merlin I was tired of trying to get my vocal for a song, i went for something else, it wasnt exactly what i wanted but suno made a good song!! Im glad

[north sleet]

I used the Essence Explorer in ChatGPT to help me define the style and to expand the words out to try to get the AI to slur or mumble the words. It only took 25-30 generations or so to find one that I thought was good enough. Of course I ended up expanding the lyrics as I progressed. I didn't add any song structure as the AI kept wanting to sing the lines instead of rapping it. This could have been user error as I think this is my first rap track with Suno.

[golden rover]

I get that often, im trying to get a sing with a metal core with a male growling but with another style completely for the chorus,

I have prompt alot of things for it, nothing works. Ai keep bleed the genre

[golden rover]

I dont understand how the expend works, i tried and most of the time i gave up

[amber vigil]

you did pretty dang good for a first attempt! Thanks for the thorough explanation too. Very much appreciated.

I keep my stuff pretty simple and have a lot to learn so all the details are helpful.

[charred merlin]

Half of the songs on my first album were experiments from me just messaging around and I just so happened to fall in love with a certain part of it for whatever reason, and I sculpted those cool parts into finished songs -- whether I just used the beginning, or I took the songs into an external DAW and clipped out a part to start a song with, sometimes you accidentally hit gold while digging for manure 😛

[leaden lark]

That was me when I did my first atmospheric pop song, was thinking of something else and since then, I’m known as the chill pop person

[charred merlin]

@golden rover — it makes me really happy that you reached out to share your journey trying to tackle those vocals (evil ones, right? SixStringToker shared a really creepy vocal intro, perhaps you could ask them to extend from it to “lock in” the vocals?)

[charred merlin]

New pfp really pops!

[leaden lark]

Did this one today

[golden rover]

i was hoping for a hard nu-metal with some jpop chorus, AI blended it again but end up with something good.
i realized that my prompt wasnt clear enough(i was testing)
prompt: Nu-metal, male screamed/clean vocals, [chorus: female J-pop style], smooth transition, return to Nu-metal.
and turned into this....
https://suno.com/song/db659729-90bd-4108-983d-b46757d8702d

so what i understand now, i need to extend, a the time Stamp b4 my Chorus, Prompt for a (jpop song/transition), merge em and Extend it at the end of the chorus to promp for a NU-metal song etc ? cuz i would like to end up making a song with some transition of genre mid song.. XD

[golden rover]

its only the image

[charred merlin]

We were working on ai image stuff in #🗣️┃chill-chat , it’s her new profile picture 🙂

[golden rover]

oh ok, cool, it's really Nice btw

i need a cool profil pic too now XD

[leaden lark]

I mean I could make one

[rare oak]

Can't generate songs tonight (bad voices quality), general problem or my lyrics may be cursed again ?

[golden rover]

weird, maybe you have Cap * () ' " Bad [ ] in the lyrics ?

[rare oak]

Problem found, ultra stereo kicks may cause problems